Herdianti Darwis

Full Author List

Purnawansyah; Syafie, Lukman; Herdianti, Universitas Muslim Indonesia

Abstract

Shallots are one of the most widely produced crops in Enrekang Regency. The obstacle in cultivation is the presence of disease in the plant which can reduce production yields. We can recognize this disease from the spots on the leaves because these spots have unique color and texture characteristics. The aim of this research is to determine the results of the classification of shallot plant diseases which focuses on purple spot and moler disease. The classification algorithms used are CNN and SVM with RBF, linear, sigmoid and polynomial kernels. The feature extraction method used is Gray Level Co-occurance Matrix (GLCM). The analysis was carried out using 320 datasets with 2 classes, namely, purple spot disease and moler disease, each class has 160 datasets. The test results show that the CNN and SVM methods with RBF, linear and polynomial kernels get accuracy, precision, recall and F1 scores of 100% respectively. Meanwhile, the SVM method on the sigmoid kernel using texture feature extraction with the GLCM method states that the accuracy value is 75%, precision 75%, recall 73% and F1-Score 74%. So these results state that the Sigmoid method using GLCM feature extraction has the lowest value among other methods

Related SDGs Targets

Full Author List

Purnawansyah; Ali, Zahrizhal; Darwis, Herdianti; Ilmawan, Lutfi Budi; Jabir, Sitti Rahmah; Manga, Abdul Rachman, Department of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

Code Large Language Models (Code LLMs) such as Code LLaMa and StarCoder have exhibited outstanding proficiency in tasks required for specific tasks like code generation. Several conducted research to similar task by utilizing fine-tuning techniques from state-of-the-art base models for more specific related task. However, due to the cost limitations and limited computing resources, performing fine-tuning from large language models is excessively high. In this study, we utilized Low-Rank Adaptation (LoRA) for base large language models such as LLaMA-2 and Phi-1.5, which uses trainable rank decomposition matrices. Furthermore, we injected Quantized LoRA (QLoRA) to help reduce memory usage while training the model and analyzed the contribution to GPU usage. Notably, our findings reveal that employing these techniques for fine-tuning on small datasets yields cost-effective and viable alternatives for language-related tasks, showcasing competitive performance compared to state-of-the-art models like CodeLLaMa 7B substantiated by lower train loss achieved in our experiments.

Related SDGs Targets

Full Author List

Purnawansyah; Hasanuddin, Tasrif; Herdianti, Universitas Muslim Indonesia

Abstract

Indonesia is a rich country in herbal plants that can be used as traditional medicine. Leaves are one of the main components of herbal plants that are difficult to distinguish in texture and shape. This study aims to classify two types of herbal leaves, namely Sauropus androgynus and Moringa leaves using the K-nearest neighbor (KNN) and Support vector machine (SVM) with fourier descriptor (FD) feature extraction on texture and shape features. The research uses primary data collected through a smartphone camera as much as 480 image data with light and dark scenarios which are then divided into 80:20 training and testing data. Based on the research that has been done, it is found that the KNN for light scenario data and dark scenarios get 92% and 94% accuracy respectively. The test results using SVM with FD feature extraction obtain an accuracy of 96% for light and dark scenarios. Thus, SVM is more recommended in the classification of herbal leaf images.

Related SDGs Targets

Full Author List

Umar, Fitriani; Darwis, Herdianti; Universitas Muslim Indonesia, Indonesia

Abstract

Watermarking citra digital banyak digunakan dan diteliti untuk masalah kepemilikan (ownership), proteksi, autentikasi dan identifikasi pemilik data digital. Skema watermarking yang baik yaitu memenuhi kriteria imperceptibility, robustness dan capacity. Penelitian ini bertujuan untuk membuat skema watermarking yang memiliki imperceptibility yang baik dan tahan (robust) terhadap serangan menggunakan Stationary Wavelet Transform (SWT). Pengujian dilakukan pada dua level transformasi dengan serangan Salt and Pepper, Speckle, Gaussian Noise, Blur, kompresi dan rotasi untuk robustness. Metode yang digunakan menghasilkan impercebtility yang baik dengan nilai PSNR di atas 70 dB pada level 1 dan di atas 40 dB pada level 2. Adapun pengujian robustness juga menunjukkan hasil yang sangat baik dengan nilai Normalized Correction di atas 0.9466 di level 1 dan di atas 0.9714 pada level 2. Ini juga menunjukkan bahwa dari segi imperceptibility, nilai PSNR padalevel 1 lebih baik dibandingkan pada level 2. Dari segi robustness, level 2 lebih baik daripada level 1.

Related SDGs Targets

Full Author List

Hardianti, Arisa Tien; Manga, Abdul Rachman; Darwis, Herdianti, Universitas Muslim Indonesia Fakultas Ilmu Komputer Makassar, Indonesia

Abstract

Jurnal merupakan sebuah dokumen yang membahas mengenai sebuah penelitian dan berfokus pada 1 bidang kelimuan. Dalam jurnal keilmuan teknik informatika, ada banyak kategori yang bisa ditentukan, semakin banyak jurnal yang terbit, semakin banyak bidang keilmuan yang dapat dikelompokkan. Dalam penelitian ini, metode naïve bayes digunakan untuk mengklasifikasikan judul jurnal berdasarkan bidang keilmuan dalam dunia ilmu komputer yang terdapat pada data jurnal Seminar Nasional Ilmu Komputer (SNRIK) tahun 2016 dengan menghitung akurasi data. Selain itu, digunakan stopword removal untuk menghilangkan kata yang tidak memiliki arti, stemming untuk mendapatkan kata dasar dari judul jurnal yang akan dikelompokkan serta inverted index yaitu matriks antara term dan data. Dari penelitian yang dilakukan menghasilkan akurasi data sebesar 50% sehingga judul jurnal mendapatkan kategori yang sesuai.

Related SDGs Targets

Full Author List

Zahirah, Dinna; Purnawansyah; Kurniati, Nia; Darwis, Herdianti, Informatics Engineering, Faculty of Computer Science, Universitas Muslim Indonesia

Abstract

Geographical position and having a tropical climate make Indonesia known for its abundant biodiversity, one of which is herbal leaves. Indonesia has more than 2039 species that fall into the category of herbal medicinal plants. Herbal leaves are plants that are used as an alternative to natural disease healing. The large number of herbal leaf plants makes it difficult for people to distinguish between herbal plants and non-herbal plants, except when herbal leaf plants bear fruit or bloom. With advances in technology, many studies have been conducted to identify types of herbal plants, one of which is to identify the characteristics of the leaves. In this study, image recognition of herbal leaves was carried out using the K-Nearest Neighbor and Convolutional Neural Network methods with feature extraction of the Gray Level Co-occurance Matrix. By using these 2 methods, the data collected in this study were 480 leaf images which were then divided into 80% testing data and 20% training data. The data used are in the form of Sauropus androgynus and Moringa leaves. Based on the test results, the Convolutional Neural Network method which is suggested in the herbal leaf image classification which has an accuracy value of 96%.

Related SDGs Targets

Full Author List

Darwis, Herdianti; Puspitasari, Rahma; Purnawansyah; Astuti, Wistiani; Atmajaya, Dedy; Hasnawi, Mardiyyah, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

This research investigates the use of deep learning for automatic waste classification, specifically using ResNet50 for feature extraction and combining it with various classification algorithms. The dataset comprises 1889 images categorized into four classes: plastic, metal, cardboard, and paper. Two approaches were evaluated: direct classification and feature extraction with ResNet50. The direct classification models, including SVM, KNN, and Random Forest, resulted in low performance, with an average accuracy of 60%. However, using ResNet50 for feature extraction significantly improved the classification accuracy across all models, with the combination of ResNet50 and SVM achieving an accuracy of 91%, and precision, recall, and F1-Score exceeding 92%. This demonstrates the effectiveness of ResNet50's feature extraction capability in enhancing the classification of images. The findings suggest that combining feature extraction and classification models provides a more accurate and efficient solution for automatic waste management systems, supporting the recycling process and waste management efficiency.

Related SDGs Targets

Full Author List

Purnawansyah; Herdianti; Satra, Ramdan, Universitas Muslim Indonesia

Abstract

Tanaman bawang merah merupakan salah satu tanaman penting dalam industri pertanian. Penyakit pada tanaman bawang merah dapat mengakibatkan kerugian yang signifikan bagi petani dan produsen. Penelitian ini bertujuan untuk mengklasifikasikan penyakit bawang merah pada daun bawang merah yang disebabkan oleh bercak ungu dan moler. Pengumpulan data citra bawang merah dilakukan secara langsung yang dilanjutkan dengan tahap pre-processing sebelum pengklasifikasian penyakit pada tanaman bawang merah. Algoritma Naive Bayes dan CNN dengan ekstraksi fitur GLCM digunakan dalam penelitian ini untuk melakukan perbandingan klasifikasi antara dua metode tersebut dalam mengklasifikasikan penyakit tanaman bawang merah yaitu bercak ungu dan moler. Hasil pengujian dengan menggunakan citra sebanyak 160 penyakit moler dan 160 penyakit bercak ungu menunjukkan bahwa kedua algoritma klasifikasi Naive Bayes dan CNN dengan ekstraksi fitur GLCM mampu mengklasifikasikan penyakit moler dan penyakit bercak ungu pada daun bawang merah dengan akurasi yang baik sebesar 100%.

Related SDGs Targets

Full Author List

Herdianti; Anraeni, Siska, Universitas Muslim Indonesia

Abstract

Peningkatan penggunaan antibiotik secara global termasuk di Indonesia, seringkali irasional dan tanpa resep berpotensi menyebabkan resistensi bakteri. Analisis sentimen data Twitter menggunakan query "antibiotik" dapat membantu mengungkap opini publik. Penelitian ini bertujuan untuk menerapkan algoritma Support Vector Machine (SVM) dengan kernel linear, RBF, dan polynomial, menggabungkan berbagai metode seperti pelabelan dengan RoBERTa, pelatihan dengan 5 cross validation, dan tokenizing bigram. Tiga skenario digunakan dalam penelitian ini dan yang menghasilkan nilai akurasi tertinggi yaitu skenario ketiga yang menggunakan slangword dari ramaprakoso dan stopword dari sastrawi sebagai refrensi library untuk filtering, nilai setiap kernel: akurasi 99,88%, presisi 99,88%, recall 99,88%, dan f1 score 99,88%. Metode SMOTE juga mempengaruhi hasil ini. Dari hasil pengujian, dapat disimpulkan bahwa SVM efektif untuk analisis sentimen.

Related SDGs Targets

Full Author List

Purnawansyah, Department of Information Systems, Universitas Muslim Indonesia, Indonesia; Wibawa, Aji Prasetya; Widiyaningtyas, Triyanna Department of Electrical Engineering and Informatics, Universitas Negeri Malang, Indonesia; Haviluddin, Department of Informatics, Universitas Mulawarman, Indonesia; Raja, Roesman Ridwan; Darwis, Herdianti, Department of Informatics, Universitas Muslim Indonesia, Indonesia; Nafalski, Andrew, UniSA Education Futures, School of Engineering, University of South Australia, Australia

Abstract

Network congestion arises from factors like bandwidth misallocation and increased node density leading to issues such as reduced packet delivery ratios and energy efficiency, increased packet loss and delay, and diminished Quality of Service and Quality of Experience. This study highlights the potential of deep learning and ensemble learning for network congestion analysis, which has been less explored compared to packet-loss based, delay-based, hybrid-based, and machine learning approaches, offering opportunities for advancement through parameter tuning, data labeling, architecture simulation, and activation function experiments, despite challenges posed by the scarcity of labeled data due to the high costs, time, computational resources, and human effort required for labeling. In this paper, we investigate network congestion prediction using deep learning and observe the results individually, as well as analyze ensemble learning outcomes using majority voting, from data that we recorded and clustered using K-Means. We leverage deep learning models including BPNN, CNN, LSTM, and hybrid LSTM-CNN architectures on 12 scenarios formed out of the combination of level datasets, normalization techniques, and number of recommended clusters and the results reveal that ensemble methods, particularly those integrating LSTM and CNN models (LSTM-CNN), consistently outperform individual deep learning models, demonstrating higher accuracy and stability across diverse datasets. Besides that, it is preferably recommended to use the QoS level dataset and the combinations of 3 clusters due to the most consistent evaluation results across different configurations and normalization strategies. The ensemble learning evaluation results show consistently high performance across various metrics, with accuracy, Matthews Correlation Coefficient, and Cohen's Kappa values nearing 100%, indicates excellent predictive capability and agreement. Hamming Loss remains minimal highlighting the low misclassification rates. Notably, this study advances predictive modeling in network management, offering strategies to enhance network efficiency and reliability amidst escalating traffic demands for more sustainable network operations.

Related SDGs Targets

Full Author List

Magfirah; Hayati, Lilis Nur; Darwis, Herdianti, Fakultas Ilmu Komputer, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

LMS AFDAL is a learning system that is applied within the scope of SMPIT Al-Fityan School Gowa. This LMS is used to facilitate modern learning. Evaluation of the AFDAL LMS is the first step to assess whether the LMS is well received or not by users. There are many approaches that can be taken in evaluating, one of which is usability evaluation. This study aims to determine the level of usability based on the System Usability Scale method with five variables, namely learnability, efficiency, memorability, errors, and satisfaction with 10 statements as a measure of quality in terms of the usability of the LMS. This research was conducted by distributing questionnaires using google form to 306 respondents consisting of teachers and students via whatsapp. Data processing uses IBM SPSS V26 and Microsoft Excel 2019. The results of the validity and reliability tests are declared valid and reliable. The results of the SUS test show that the final SUS value of 306 respondents' responses is 64.6, according to the rules of SUS interpretation that the score is 64.6 for the Acceptability Ranges level, namely Marginal (quite acceptable), the Grade Scale results in terms of user acceptance levels are included in the C- level, and Adjectives The rating is included in the OK category. These results indicate that the AFDAL LMS is quite accepted by its users, but this figure is quite low so that some improvements are needed to make it even better.

Related SDGs Targets

Full Author List

Rahmadani, Aulia Rezky; Purnawansyah; Darwis, Herdianti; Ilmawan, Lutfi Budi, Program Studi Teknik Informatika, Fakultas Ilmu Komputer, Universitas Muslim Indonesia,

Abstract

Daun merupakan salah satu komponen tumbuhan yang mengandung khasiat alami dan bermanfaat untuk menjaga kesehatan manusia. Namun beberapa jenis daun memiliki ciri dan karakteristik yang sama sehingga sulit untuk dibedakan. Penelitian ini bertujuan untuk melakukan pengklasifikasian jenis daun herbal dengan menggunakan metode SVM dengan empat kernel (Linear, RBF, Polynomial, Sigmoid) dan CNN dengan ekstraksi fitur Fourier descriptor (FD). Dataset yang diolah adalah citra daun katuk, dan daun kelor sejumlah 480 citra yang terbagi menjadi data training dan data testing dengan perbandingan 80%:20%, 70%:30% dan 60%:40%menggunakan dua skenario yaitu gelap dan terang. Dariproses pengujian diperolehhasil pengujian algoritma FD + SVM memberikan hasil yang lebih baik memperoleh nilai accuracy pada kernel linear, RBF, polynomial, sebesar 100% pada perbandingan 80%:20%, 70%:30% dan 60%:40%pada skenario terang dan gelap. Sedangkan hasil dari proses pengujian FD + CNN pada perbandingan 70%:30% pada skenario terang mendapatkan nilai accuracy sebesar 100%. Dengan demikian, algoritma FD + CNN pada skenario terang dan algoritma FD + SVM dengan kernel Linear, RBF, polynomial dapat direkomendasikan dalam pengklasifikasian citra daun herbal.

Related SDGs Targets

Full Author List

Rahmadani, Aulia Rezky; Purnawansyah; Darwis, Herdianti; Ilmawan, Lutfi Budi, Program Studi Teknik Informatika, Fakultas Ilmu Komputer, Universitas Muslim Indonesia

Abstract

Leaves are one component of plants that contain natural properties and are useful for maintaining human health. However, several types of leaves have the same characteristics and characteristics that make it difficult to distinguish. This study aims to classify types of herbal leaves using the SVM method with four kernels (Linear, RBF, Polynomial, Sigmoid) and CNN with Fourier descriptor (FD) feature extraction. The processed dataset is katuk leaf images, and Moringa leaf images of 480 images which are divided into 80% training data and 20% testing data using two scenarios, namely dark and light. From the testing process, it was found that FD + CNN in the light and dark scenarios obtained an accuracy value of 98%. Thus, the FD + SVM algorithm with Linear, RBF, polynomial kernels can be recommended in classifying herbal leaf images to have the best accuracy value of 100%.

Related SDGs Targets

Full Author List

Salim, Yulita; Rakasyah, Athar Fathana; Darwis, Herdianti; Herlinda; Irawati; Manga, Abdul Rachman, Departement of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

In the era of burgeoning image data, the demand for robust image classification algorithms has never been more pressing. This research article delves into the realm of image classification, encompassing a comprehensive analysis of diverse datasets and machine learning algorithms, While simultaneously studying the efficiency of ensemble approaches in improving classification performance. The study employs five distinct datasets, spanning medical images, everyday objects, and natural scenery, ensuring a broad spectrum of classification challenges. These datasets include the Skin Cancer ISIC dataset, CIFAR-10, Flowers, Apparel Image, and Brain Tumor dataset. The image classes within these data sets vary significantly, presenting an ideal testbed for assessing algorithmic versatility. Our investigation scrutinizes five machine learning algorithms: Support Vector Classifier (SVC), Random Forest Classifier (RFC), Gradient Boosting Classifier (GBC), K-Nearest Neighbors (KNN), and Gaussian Naive Bayes (GNB). Each algorithm is evaluated individually, followed by a comprehensive ensemble approach involving a Voting Classifier. The results unveil nuanced performance variations across diverse datasets. Notably, RFC and GBC exhibit remarkable accuracy in brain tumor image classification, while KNN demonstrates strengths in classifying apparel images. Ensemble techniques, embodied by the Voting Classifier, harmonize these algorithms, yielding competitive and balanced performance across the datasets. This article contributes valuable insights into the realm of image classification, shedding light on algorithmic strengths and limitations, the efficacy of ensemble techniques, and their applicability to diverse image datasets. These findings hold significance for fields ranging from medical diagnostics to everyday object recognition, paving the way for more precise and versatile image classification solution

Related SDGs Targets

Full Author List

Darwis, Herdianti; Pagala, Arya Nanda Pratama; Anraeni, Siska; Amaliah, Tazkirah; As'ad, Ihwana; Tenripada, Andi Ulfah, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

Childfree is the choice to live without having or adopting children. This phenomenon is still considered a sensitive topic in Indonesia, where the prevailing belief holds that the primary goal of marriage is to have children. Numerous individuals share their perspectives on this matter through Twitter. In this research, 6654 raw data have been collected from Twitter using crawling techniques using Rapidminer and scraping using website scrape based on the keyword “childfree” which is preprocessed into clean data. The sentiment analysis model carried out includes categorizing childfree sentiment based on religious, medical and economic fields, then measuring the performance of the support vector machine, involving several methods such as RoBERTa labeling, bigram tokenizing, term frequency inverse document frequency weighting, 5 cross validation training, and synthetic minority over-sampling technique with edited nearest neighbors. The research results show that the economic category is the most influential field with 722 related sentiments, the accuracy performance of SVM gives a value of 75.95% on the linear kernel and the application of SMOTEENN gives a value of 95.94% on the linear kernel, it is proven that using SMOOTEENN can overcome data imbalance.

Related SDGs Targets

Full Author List

Aszahrah, Hilma; Anraenia, Siska; Darwisa, Herdianti; Universitas Muslim Indonesia, Indonesia

Abstract

Pengenalan jenis kayu dapat dilakukan dengan melakukan pengamatan berdasarkan pada ciri-ciri tertentu yaitu dengan menggunakan parameter tekstur, berat, warna dan lain sebagainya. Begitupun dengan memilih jenis kayu sebagai bahan furniture, selama ini sering dilakukan adalah dengan melihat saja atau memegang kayu secara umum dengen memperhatikan teksturnya. Selain akurasi yang kurang, cara ini juga membutuhkan pengalaman yang cukup banyak apalagi dalam memilih kayu yang akan digunakan dalam membuat dekorasi rumah seperti jendela dan kursi dibutuhkan kayu yang kuat dan kokoh karena itu sistem pengenalan jenis kayu yang akurat dan praktis sangat penting untuk dikembangkan. Mengingat banyaknya jenis kayu yang memiliki kesamaan ciri sehingga sulit mengindetifikasi jenis kayu yang akan digunakan sebagai bahan furniture oleh karena itu dikembangkan suatu sistem teknik klasifikasi untuk identifikasi jenis kayu dengan metode GLCM. Metode ini merupakan suatu metode yang melakukan analisis terhadap suatu piksel pada citra dan mengetahui tingkat keabuan yang sering terjadi, metode yang paling sering digunakan untuk analisis tekstur yang didasarkan pada ciri statistik citra yang dimana dalam perhitungan statistiknya menggunakan distribusi derajat keabuan (histogram) dengan mengukur tingkat kekontrasan, granularitas, dan kekasaran suatu daerah dari hubungan ketetanggaan antar piksel di dalam citra. Hasil penelitian menunjukkan dari tahap pengujian dengan nilai K=1,K=7,K=10,K=12, dan K=15 dengan jumlah data training sebanyak 45 data didapatkan nilai K=1 sebagai nilai persentase tertinggi dengan tingkat akurasi sistem sebesar 91%.

Related SDGs Targets

Full Author List

Balla, Masita; Harlinda, Harlinda; Darwis, Herdianti; Universitas Muslim Indonesia, Indonesia

Abstract

Penyakit hipertensi dapat menyerang siapa saja dari berbagai kelompok umur maupun kelompok sosial ekonomi. Penentuan menu makanan dibutuhkan oleh penderita hipertensi untuk keadaan yang lebih baik. Namun sebagian besar penderita hipertensi tidak terlalu memahami asupan makanan yang harus dikonsumsi dan bagaimana pola makan yang baik bagi penderita hipertensi. Penelitian ini bertujuan untuk merancang sistem pendukung keputusan menu makanan untuk memberikan rekomendasi menu makanan yang baik untuk dikonsumsi bagi penderita hipertensi. Dalam proses pengolahan data untuk memudahkan pengambilan keputusan dalam penelitian ini menggunakan metode Wighted Aggregated Sum Product Assesment (WASPAS) yang akan melakukan proses perankingan berdasarkan kriteria dan bobotnya untuk menentukan alternatif mana yang lebih optimal untuk pemilihan menu makanan bagi penderita hipertensi. Sistem pendukung keputusan ini dibangun secara optimal dengan tingkat akurasi sebesar 100% dengan menggunakan sampel data sebanyak 40 jenis alternatif menu makanan dan tiga kriteria yaitu lemak, protein dan natrium.

Related SDGs Targets

Full Author List

Balla, Masita; Harlinda, Harlinda; Darwis, Herdianti; Universitas Muslim Indonesia, Indonesia

Abstract

Penyakit hipertensi dapat menyerang siapa saja dari berbagai kelompok umur maupun kelompok sosial ekonomi. Penentuan menu makanan dibutuhkan oleh penderita hipertensi untuk keadaan yang lebih baik. Namun sebagian besar penderita hipertensi tidak terlalu memahami asupan makanan yang harus dikonsumsi dan bagaimana pola makan yang baik bagi penderita hipertensi. Penelitian ini bertujuan untuk merancang sistem pendukung keputusan menu makanan untuk memberikan rekomendasi menu makanan yang baik untuk dikonsumsi bagi penderita hipertensi. Dalam proses pengolahan data untuk memudahkan pengambilan keputusan dalam penelitian ini menggunakan metode Wighted Aggregated Sum Product Assesment (WASPAS) yang akan melakukan proses perankingan berdasarkan kriteria dan bobotnya untuk menentukan alternatif mana yang lebih optimal untuk pemilihan menu makanan bagi penderita hipertensi. Sistem pendukung keputusan ini dibangun secara optimal dengan tingkat akurasi sebesar 100% dengan menggunakan sampel data sebanyak 40 jenis alternatif menu makanan dan tiga kriteria yaitu lemak, protein dan natrium.

Related SDGs Targets

Full Author List

Purnawansyah; Herdianti; Harlinda, Universitas Muslim Indonesia

Abstract

Bawang merah merupakan suatu kebutuhan masyarakat terutama pada bahan makanan dan juga digunakan untuk Kesehatan. Dengan banyaknya manfaat bawang merah, dibalik itu juga memiliki suatu kendala atau resiko pada penanaman bawang merah salah satu resikonya adalah hama atau penyakit yang dapat merugikan petani bawang merah. Tujuan dari penelitian ini yaitu mengklasifikasi penyakit daun bercak ungu dan moler pada tanaman bawang merah, yang di implementasikan menggunakan metode ekstraksi fitur Gray Level Co-Occurance Matix (GLCM) yang digunakan untuk ekstraksi fitur tekstrur. Selain itu ada lima jarak yaitu Eucludiean, Manhattan, Chebyshev, Minkowski, Hamming digunakan dalam metode klasifikasi K-Nearest Neighbor (KNN). Penelitian ini juga menggunakan metode klasifikasi Convolutional Neural Network (CNN). Hasil dari penelitian ini yang diperoleh menggunakan metode GLCM dan KNN dengan jarak Euclidean, Manhattan, Chebyshev, dan Minkowski mendapatkan hasil akurasi yang tinggi yakni sebesar 100%, sedangkan nilai akurasi terendah terdapat pada KNN jarak Hamming nilai akurasi yaitu sebesar 42%, adapun klasifikasi dari gabungan dari metode GLCM dan CNN mendapatkan hasil akurasi sebesar 100% dan pada metode CNN yang tanpa metode ekstraksi memiliki nilai akurasi sebesar 100%

Related SDGs Targets

Full Author List

Astuti, Wistiani, Univeristas Muslim Indonesia; Wibawa, Aji Prasetya, Universitas Negeri Malang; Haviluddin, Universitas Mulawarman; Darwis, Herdianti, Univeristas Muslim Indonesia

Abstract

One prevalent conversational system within the realm of natural language processing (NLP) is chatbots, designed to facilitate interactions between humans and machines. This study focuses on predicting frequently asked questions by students using the Duel Intent and Entity Transformer (DIET) Classifier method and assessing the performance of this method. The research involves employing 300 epochs with an 80% training data and 20% testing data split. In this study, the DIET Classifier adopts a multi-task transformer architecture to simultaneously handle classification and entity recognition tasks. Notably, it possesses the capability to integrate diverse word embeddings, such as BERT and GloVe, or pre-trained words from language models, and blend them with sparse words and n-gram character-level features in a plug-and-play manner. Throughout the training process of the DIET Classifier model, data loss and accuracy from both training and testing datasets are monitored at each epoch. The evaluation of the text classification model utilizes a confusion matrix. The accuracy results for testing the DIET Classifier method are presented through four case studies, each comprising 25 text messages and 15 corresponding chatbot responses. The obtained accuracy values range from 0.488 to 0.551, F1-Score values range from 0.427 to 0.463, and precision range from 0.417 to 0.457.

Related SDGs Targets

Full Author List

Umar, Fitriyani; Herdianti; Astuti, Wistiani, Universitas Muslim Indonesia

Abstract

Banyak upaya yang dapat dilakukan untuk membantu proses pembelajaran. Transfer pengetahuan akan optimal jika didukung dengan media pembelajaran yang tepat. Sekolah Dasar Inpres Marinding dalam proses pembelajaran selama ini belum dapat memanfaatkan teknologi terkini yang sesuai untuk meningkatkan pemahaman siswa tentang suatu mata pelajaran. Padahal pesatnya perkembangan teknologi informasi dapat dimanfaatkan sebagai media tambahan untuk mendukung pembelajaran yang merangsang imajinasi, interaktif dan menumbuhkan minat belajar sehingga proses belajar mengajar menjadi lebih baik lagi. Demi menunjang proses pembelajaran di SDN 294 Inpres Marinding, siswa diharapkan memiliki banyak buku yang berisi satu tema tertentu sebanyak jumlah tema yang ada. Akan tetapi, pembelajaran masih berpusat pada buku tersebut dan tidak ada alat peraga khusus tentang Flora. Siswa tidak dapat melihat objek Flora secara langsung, hanya melalui gambar di buku dan kurang detailnya informasi tentang objek tersebut. Akibatnya, pembelajaran cenderung monoton, dan kurang kreatifitas. Solusi yang diusulkan adalah memberikan pelatihan untuk implementasi aplikasi Augmented Reality Pengenalan dengan tujuan untuk meningkatkan pengetahuan guru dan siswa dalam pemanfaatan teknologi informasi untuk media pembelajaran alternatif melalui Augmented Reality khususnya pembelajaran Flora. Tercapainya tujuan kegiatan telah menghasilkan luaran berupa modul dan aplikasi yang dapat menjadi alternatif media pembelajaran, publikasi pada media online dan jurnal yang diterbitkan di ILKOMAS

Related SDGs Targets

Full Author List

Purnawansyah; Herdianti; Astuti, Wistiani, Universitas Muslim Indonesia

Abstract

Tumbuhan herbal memiliki banyak variasi yang dapat dikenali melalui ciri uniknya secara visual. Namun, cara ini sulit diterapkan pada tumbuhan yang memiliki ciri hampir sama. Penelitian ini membandingkan kinerja metode K-Nearest Neighbour (KNN) dan Convolutional Neural Network (CNN) dalam klasifikasi fitur daun herbal yang diekstraksi dengan menggunakan Geometric Moment Invariant (GMI) dan Hue Saturation Value (HSV). Dataset yang digunakan adalah dataset citra daun katuk (Sauropus androgynus) dan daun kelor (Moringa oleifera) dengan skenario citra terang dan citra gelap. Pembagian data untuk tiap skenario adalah 80% untuk training dan 20% untuk testing. Metode KNN diuji menggunakan nilai dan evaluasi kinerja KNN dan CNN meliputi accuracy, precision, recall, dan f1-score. Hasil penelitian menunjukkan bahwa CNN tanpa ekstraksi fitur dan CNN dengan kombinasi ekstraksi fitur HSV memperoleh performa terbaik dengan rata-rata nilai precision, recall, f1-score dan accuracy sebesar 98% untuk skenario gelap maupun terang

Related SDGs Targets

Full Author List

Purnawansyah; Darwis, Herdianti, Universitas Muslim Indonesia, Makassar, Indonesia; Raja, Roesman Ridwan, Kyushu Institute of Technology, Iizuka City, Jepang

Abstract

This research explores public sentiment towards the Indonesian police using sentiment analysis and machine learning techniques. The study addresses the challenge of understanding public opinion based on social media comments related to significant police cases. The aim is to compare reported satisfaction levels with actual public sentiment. Utilizing the Indonesian RoBERTa base IndoLEM sentiment classifier, comments were analyzed and preprocessed. The classification was conducted using Random Forest (RF) and Complement Naive Bayes (CNB) models, incorporating unigram and bi-gram features. Oversampling techniques were applied to handle data imbalance. The best-performing model, Random Forest with bi-gram features, achieved high evaluation scores, including a precision of 0.91 and accuracy of 0.91. The findings reveal significant insights into public opinion, contributing to improved law enforcement strategies and public trust.

Related SDGs Targets

Full Author List

Triyadi, Azwan; Purnawansyah; Darwis, Herdianti, Universitas Muslim Indonesia

Abstract

Elon Musk owns the business Neuralink, which attempts to build brain-machine interfaces. This study categorizes public opinion towards the use of Neuralink goods, including whether people agree (positive), disagree (negative), or feel neither way. Without accessing the Twitter API, the Twint Python Libraries were utilised to retrieve a dataset of 3000 using the keyword “neuralink”. What datasets are included in positive, neutral, or negative categories are designated using RoBERTa. Term Frequency Inverse Document Frequency (TF-IDF) is utilized for feature extraction, while Synthetic Minority Over-sampling Technique (SMOTE) is employed to handle class imbalance. Complement Naive Bayes, achieved accuracy of 81%, followed by Multinomial Naive Bayes, which achieved accuracy of 80%, and Gaussian Naive Bayes, which achieved accuracy of 75%. The model Complement Naïve Bayes was used in this study to attain the maximum accuracy, and accuracy increases when employing SMOTE compared to other Naïve bayes variants.

Related SDGs Targets

Full Author List

Manga, Abdul Rachman; Nirmala; Azis, Huzain; Fattah, Farniwati; Salim, Yulita; Darwis, Herdianti, Departement of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

Flower image classification poses a challenge in digital image processing, requiring effective methods for feature extraction and classification. The aim of this research is to improve the accuracy of flower image classification by employing ResNet-50 for feature extraction, with and without segmentation, and evaluating the effectiveness of various classification algorithms. The dataset consists of images of Calendula and Coreopsis flowers, totaling 2,025 samples. Four segmentation techniques-Canny, Thresholding, Otsu, and Mean Shift-along with a non-segmentation approach are applied. The features extracted using ResNet-50 are classified with Naive Bayes, Support Vector Machine (SVM), Decision Tree, and K-Neighbors Classifier (KNN). Performance evaluation is conducted using accuracy, precision, recall, and F1-score with 5-fold cross-validation. The results show that SVM delivers the best performance in most scenarios. The highest accuracy in segmentation scenarios was achieved with the Mean Shift technique at 0.91, while the non-segmentation approach yielded the highest accuracy of 0.95. The non-segmentation approach proves more effective, indicating segmentation is not always required for high classification accuracy with ResNet-50. This study shows that ResNet-50, especially without segmentation, can significantly improve flower image classification compared to traditional methods, opening opportunities for more efficient systems.

Related SDGs Targets

Full Author List

Satra, Ramdan; Dahlan, Imram Afdillah; Darwis, Herdianti; Purnawansyah; Mujaddid, Syariful; Fattah, Farniwati, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

This study focuses on comparing the performance of K-Nearest Neighbors (KNN), TabNet, and Wide & deep learning methods in classifying Distributed Denial of Service (DDoS) attacks on Software-Defined Networks (SDN). The use of SDN enables centralized control of network infrastructure, making it vulnerable to DDoS attacks occur due to the centralized nature of SDN architecture. Machine learning models, including KNN, TabNet, and Wide & deep learning, are applied to an SDN-specific DDoS dataset to evaluate their effectiveness in accurately classifying normal and malicious traffic. These models were tested using various data splits (60:40, 70:30, 80:20, and 90:10) to determine the optimal ratio for training and validation. KNN exhibited the highest accuracy, reaching 98% in both 80:20 and 90:10 splits, while wide & deep learning achieved 94.99% accuracy, and TabNet demonstrated a 93.59% accuracy. The results suggest that KNN, despite being a simpler algorithm, outperforms the more complex deep learning models in this specific task. The findings provide valuable insights for researchers and network administrators in selecting effective machine learning algorithms for DDoS detection in SDN environments.

Related SDGs Targets

Full Author List

Purnawansyah; Adnan, Adam; Darwis, Herdianti, Faculty of Computer Science, Universitas Muslim Indonesia, Jl. Urip Sumoharjo KM 5, Makassar, 90231, Indonesia; Wibawa, Aji Prasetya, Universitas Negeri Malang, Jl. Semarang No. 5, Malang, 65145, Indonesia; Widyaningtyas, Triyanna; Haviluddin, Universitas Mulawarman, Jl. Kuaro, Samarinda, 75119, Indonesia

Abstract

Facial Expression Recognition (FER) plays a crucial role in humancomputer interaction, yet improving its accuracy remains a significant challenge. This study aims to enhance the robustness and effectiveness of FER systems by integrating multiple machine learning techniques within a semi-supervised learning framework. The primary objective is to develop a more effective ensemble model that combines Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), Support Vector Classifier (SVC), and Random Forest classifiers, utilizing both labeled and unlabeled data. The research implements data augmentation and feature extraction techniques, utilizing advanced architectures such as VGG19, ResNet50, and InceptionV3 to improve the quality and representation of facial expression data. Evaluations were conducted across three dataset scenarios: original, feature-extracted, and augmented, using various labelto- unlabeled ratios. The results indicate that the ensemble model achieved a notable accuracy improvement of 87% on the augmented dataset compared to individual classifiers and other ensemble methods, demonstrating superior performance in handling occlusions and diverse data conditions. However, several limitations exist. The study's reliance on the JAFFE dataset may restrict its generalizability, as it may not cover the full range of facial expressions encountered in real-world scenarios. Additionally, the effect of label-to-unlabeled ratios on the model's performance requires further exploration. Computational efficiency and training time were also not evaluated, which are critical considerations for practical implementation. For future research, it is recommended to employ cross-validation methods for more robust performance evaluation, explore additional data augmentation techniques, optimize ensemble configurations, and address the computational efficiency of the model to better advance FER technologies.

Related SDGs Targets

Full Author List

Hayati, Lilis Nur; Randana, Fitrah Yusti; Darwis, Herdianti, Universitas Muslim Indonesia

Abstract

Machine learning-based sentiment analysis has become essential for understanding public perceptions of public services, including air transportation. Sultan Hasanuddin Airport, one of the main gateways in eastern Indonesia, faces the challenge of improving services amid changing user needs due to the COVID-19 pandemic. This study aims to compare the effectiveness of three machine learning algorithms- Support Vector Machine (SVM), Naive Bayes Multinomial, and K-Nearest Neighbor (KNN)-in analyzing the sentiment of user reviews related to airport services. The research also explores data splitting techniques, text preprocessing, data balancing using SMOTE, model validation, and method parameterization to ensure optimal results. The review data was retrieved from Google Maps (2021-2024) and underwent manual labelling. Text preprocessing includes normalization, stemming using Sastrawi, and stopword removal. The data-balancing technique uses SMOTE, while model evaluation is done with stratified k-fold cross-validation. SVM with a linear kernel showed the best performance, achieving an F1-score of 98.4%. Naive Bayes performed optimally, achieving an F1-score of 93.9%, while KNN recorded the best F1-score of 92.0%. SMOTE was shown to improve Naive Bayes' performance on unbalanced datasets, although it did not significantly impact SVM. The findings of this study provide data-driven recommendations to improve services at Sultan Hasanuddin Airport, such as the management of cleaning facilities, waiting room comfort, and passenger flow efficiency. In addition, this research opens up opportunities for developing real-time sentiment analysis systems that can be applied in other air transportation sectors.

Related SDGs Targets

Full Author List

Wati, Silmi Nur Zaskia; Herman; Darwis, Herdianti, Universitas Muslim Indonesia

Abstract

Pelaksanaan perkuliahan daring menggunakan KALAM di UMI banyak menuai kontroversi dikalangan mahasiswa. Banyak pendapat mahasiswa yang dikeluarkan terkait metode pembelajaran daring di UMI. Penelitian ini bertujuan menganalisis sentimen mahasiswa terkait perkuliahan daring di UMI dengan menggunakan algoritma Naïve Bayes Classifier dan K-Nearest Neighbor, serta menggabungkan berbagai metode seperti pelabelan dengan NLTK, pengujian dengan 5 cross validation, dan menggunakan unigram tokenizing. Beberapa teknik pelabelan digunakan pada penelitian ini dan menghasilkan tingkat keakuratan paling tinggi adalah pelabelan menggunakan NLTK dengan algoritma KNN dengan menggunakan SMOTE menghasilkan akurasi sebesar 100% dibandingkan dengan algoritma Naïve Bayes Classifier yang memiliki nilai akurasi sebesar 98.33%. Sehingga algoritma KNN dapat digunakan dengan baik pada pengklasifikasian sentimen mahasiswa terhadap perkuliahan daring di UMI.

Related SDGs Targets

Full Author List

Belluano, Poetri Lestari Lokapitasari; Darwis, Herdianti, Universitas Muslim Indonesia

Abstract

Bawang merah merupakan tanaman holtikultura yang berpotensi tinggi terhadap perubahan harga sehingga sangat fluktuatif bagi petani maupun konsumen dan juga termasuk komoditas trategis. Di Indonesia khusunya, pertumbuhan bawang merah mengikuti pola musim yang terjadi, sehingga pada musim tertentu stok bawang merah menurun. Prediksi harga bawang merah menjadi penting dilakukan untuk mengetahui harga bawang merah ke depan. Neural network termasuk algoritma yang terbaik dalam melakukan prediksi. Masalah utama bagaimana menentukan jumlah neuron dan hidden layer yang optimal sehingga akurasi prediksinya tinggi. Jurnal ini bertujuan untuk merancang arsitektur neural network dengan menggunakan algoritma bacpropagation. Tahapan penelitian dilakukan adalah mengumpulkan data harga bawang merah, melakukan proprecessing data, memproses prediksi, pengujian akurasi, pengujian akurasi dan eror serta implementasi. Dalam memproses prediksi dilakukan sesuai dengan rancangan model prediksi, yaitu parameter epoch, momentum, learning rate, hidden layer untuk menghasilkan keakuratan yang tinggi. Temuan yang diperoleh berupa rancangan optimal untuk melakukan prediksi yaitu dengan menggunakan multilayet. Diperoleh tingkat akurasi mencapai 98.324% atau dengan tingkat eror yang relatif rendah yaitu 11,161%

Related SDGs Targets

Full Author List

Herdianti; Manga, Abdul Rachman; Azis, Huzain, Universitas Muslim Indonesia

Abstract

Sistem informasi berbasis web telah menjadi salah satu trend pemanfaatan teknologi yang saat ini ditemui di dunia pariwisata dan ekonomi kreatif sebagai salah satu alternatif penyampaian informasi (pengetahuan dan berita) berbasis Internet yang dikeluarkan oleh dinas pariwisata atau pengelola tempat wisata dengan tujuan untuk kemudahan dalam pendistribusian informasi pada publik, baik masyarakat sekitar, wisatawan domestik, maupun wisatawan mancanegara. Sehubungan dengan hal tersebut, Fakultas Ilmu komputer sebagai lembaga yang berperan aktif dalam pengembangan dan penerapan teknologi informasi di Kawasan Indonesia Timur dengan visi fakultas yaitu “Smart Village” yang bersinergi dengan visi Universitas Muslim Indonesia terkait “Halal Issues” melakukan peningkatan kualitas dan pengembangan desa di Lembang Marinding Kecamatan Mengkendek Tana Toraja khusus di bidang pariwisata halal dan ekonomi kreatif melalui sebuah kegiatan pengabdian kepada masyarakat (PkM) yang dilaksanakan dalam bentuk pengembangan website bagi pengelola Gunung Kandora Kabupaten Tana Toraja demi peningkatan pendapatan ekonomi daerah dan secara khusus untuk kesejahteraan masyarakat sekitar

Related SDGs Targets

Full Author List

Umar, Fitriyani; Herdianti; Purnawansyah, Universitas Muslim Indonesia

Abstract

Hal yang kritis dalam proses pengenalan pola adalah ekstraksi fitur. Merupakan suatu metode untuk mendapatkan ciri-ciri suatu citra (image) sehingga dapat dikenali satu sama lain. Pada penelitian ini, metode deskriptor Fourier digunakan untuk mengekstraksi pola aksara Lontara yang terdiri dari 23 huruf. Deskriptor Fourier adalah metode yang digunakan dalam pengenalan objek dan pemrosesan citra untuk merepresentasikan bentuk batas segmen citra. Pengenalan karakter dilakukan dengan menggunakan jarak Euclidean dan Manhattan. Hasil pengujian menunjukkan bahwa tingkat pengenalan tertinggi mencapai akurasi 91,30% dengan menggunakan koefisien Fourier sebesar 50. Pengenalan huruf menggunakan Manhattan dan Euclidean cenderung sama atau menghasilkan akurasi yang cenderung serupa. Akurasi tertinggi dicapai saat menggunakan Manhattan sebesar 91,30%

Related SDGs Targets

Full Author List

Purnawansyah; Herdianti; Nurhayati, Lilis, Universitas Muslim Indonesia

Abstract

Bawang merah rentan terhadap serangan penyakit yang dapat mengganggu pertumbuhan dan mengakibatkan hasil panen yang tidak maksimal bahkan gagal panen, seperti bercak ungu dan moler. Penelitian ini bertujuan untuk mengklasifikasikan penyakit bawang merah dengan mengimplementasikan meetode naïve bayes (gaussian , bernoulli, dan multinomial) dan CNN pada citra bawang merah yang diekstraksi menggunakan fourier descriptor. Metode FD – CNN memperoleh tingkat accuracy 98% dalam mengklasifikasikan penyakut bawang merah, moler dan bercak ungu, sedangkan metode CNN tanpa menggunakan ekstraksi menghasilkan nilai accuracy sebesar 97%. Adapun pada metode naïve bayes, pengklasifikasian yang memiliki accuracy paling tinggi adalah metode gaussian naïve bayes sebesar 95% sedangkan yang paling rendah yaitu metode bernoulli naïve bayes dengan tingkat accuracy sebesar 42%. Dengan demikian, dapat disimpulkan bahwa CNN, FD-CNN, dan FD-GNB efektif untuk meningkatkan performa klasifikasi pada citra daun bawang merah.

Related SDGs Targets

Full Author List

Purnawansyah; Herdianti, Universitas Muslim Indonesia

Abstract

Google Playstore memiliki karakteristik yang berbeda dengan Apple App Store yaitu lebih terbuka terhadap developer aplikasi mobile sehingga memiliki varian yang lebih beragam dibanding dengan Apple App Store. Setiap aplikasi di dalam app store dapat dikelompokan berdasarkan karakteristik yang sama dan dapat disebut sebagai kategori dan genre. Pada tahun 2018 jumlah mobile app yang tersedia mencapai 3,6 juta aplikasi. Berbagai jenis mobile app tersedia pada layanan google play store, mulai dari hiburan, media sosial, editor, jasa transportasi, perdagangan (marketplace), dan kesehatan. Penelitian ini bertujuan untuk melakukan 5 analisis yaitu aplikasi dengan rating tertinggi, mencari 5 aplikasi dengan size paling berat (MBs), visualisasi data content ratings aplikasi, mengidentifikasi aplikasi dengan install terbanyak, visualisasi kategori aplikasi. Dari 2152981 data yang telah di crawling diperoleh bahwa 5 aplikasi dengan rating tertinggi yaitu Biliyor Musun - Sonsuz Yarış, CoronaSurveys, Amkshoproom Shopping, Merlin CRM, Tictactoe Superpowers dan free game. Fun and Chalmo, mencari 5 aplikasi dengan size paling berat (MBs) yaitu SkySafari 6 Pro, Audio Book Bible Offline Arabic, Audio Book Bible Offline Burmese, Audio Book Bible Offline Amharic dan Audio Book Bible Offline Germany, Visualisasi content data rating dari grafik dapat kita lihat bahwa mayoritas aplikasi mobile pada android mengatur content rating kedalam kategori Everyone, Mengindentifikasi aplikasi install terbanyak dari data yang telah diperoleh bahwa hanya terdapat 1 aplikasi yang memiliki jumlah install lebih dari 10M install dan 14 aplikasi yang memiliki jumlah install lebih dari 5 M, visualisasi kategori aplikasi dari data yang diperoleh bahwa aplikasi berkategori education memiliki jumlah terbanyak yang ada di pasar playstore saat ini

Related SDGs Targets

Full Author List

Rahmatullah, Muhammad Faiq; Belluano, Poetri Lestari Lokapitasari; Darwis, Herdianti, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

Google Play Store adalah salah satu platform distribusi aplikasi terbesar yang memungkinkan pengguna memberikan ulasan terhadap aplikasi yang mereka pakai. Di era digital saat ini, ulasan pengguna menjadi sumber data penting untuk menilai performa dan kualitas aplikasi. Namun, banyaknya jumlah ulasan membuat analisis secara manual menjadi kurang efisien. Oleh karena itu, peracangan ini ini mengadopsi pendekatan machine learning untuk mengklasifikasikan ulasan ke dalam kategori sentimen positif, negatif, atau netral. Proses analisis meliputi beberapa tahap, seperti pengumpulan data, praproses teks, ekstraksi fitur dengan TF-IDF, pelatihan model menggunakan Random Forest, serta evaluasi kinerja model. Hasil evaluasi menunjukkan bahwa model yang dikembangkan berhasil mengklasifikasikan sentimen dengan akurasi sebesar 68.5%, dengan performa terbaik pada sentimen negatif. Selain itu, penerapan metode Random Forest juga membuka peluang untuk pengembangan sistem analitik otomatis yang dapat digunakan oleh pengembang aplikasi dalam meningkatkan kualitas layanan mereka. Dengan memahami kecenderungan opini pengguna secara cepat dan akurat, pengambilan keputusan dalam pengembangan fitur baru atau perbaikan bug dapat dilakukan secara lebih terarah. Implementasi metode ini juga berpotensi untuk diterapkan pada sektor lain seperti e-commerce, layanan publik, atau media sosial, di mana opini pengguna menjadi salah satu aspek penting dalam evaluasi layanan

Related SDGs Targets

Full Author List

Herman; Putra Muhammad Dani Arya, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia; Nasir, Haidawati, MIIT University Kuala Lumpur, Malaysia; Darwis, Herdianti; Mansyur, St. Hajrah, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia; Noor, Megat Nurolazmi MM, MIIT University Kuala Lumpur, Malaysia

Abstract

This study focuses on being able to classify traditional Toraja carvings using Convolutional Neural Networks (CNN), focusing on three CNN architectures, specifically VGG-16, VGG-19, and a model that is concatenated from both. The aim is to determine the most effective architecture and ratio of training data and validation data sharing to achieve the highest classification accuracy. The image dataset consisting of seven different Toraja carving motifs or classes underwent data pre-processing, namely data augmentation, to improve model generalization and reduce overfitting. Experiments were conducted using four scenarios of training data and validation data separation. The final outcome of this research is that VGG-16 reached the best validation performance of 97.36% with a data 90%: 10% separation. It manifests its superior ability to Capture the information of complicated Toraja carving motifs. VGG-19 and the combined model also performed well, but the results were still below the best results of VGG-16 and emphasized that the VGG-16 architecture, especially with a data separation of 90%:10%, is the most reliable CNN architecture for accurately classifying Toraja carvings.

Related SDGs Targets

Full Author List

Herman; Darwis, Herdianti; Nurfauziyah; Puspitasari, Rahma; Widyawati, Dewi; Faradibah, Amaliah, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

This study aims to classify anxiety disorders in adolescents using three machine learning algorithms, namely naive bayes, C4.5, and K-Nearest Neighbor (K-NN). The data used was taken from the DASS-21 questionnaire, which contains 418 samples with 11 attributes, including age, gender, and seven anxiety-related questions. The algorithm was tested using the holdout technique with 80:20, 70:30, and 60:40 data splits, as well as the k-fold cross-validation technique. The results showed that the C4.5 algorithm performed best with 100% accuracy in the holdout technique, followed by naive bayes with 99% accuracy, and K-NN with 94% accuracy. In cross-validation testing, C4.5 also showed the highest accuracy of 98%, while naive bayes and K-NN achieved 89% and 92% respectively. This study concludes that the C4.5 algorithm is superior in classifying anxiety compared to naive bayes and K-NN, so it can be relied upon for machine learning-based diagnostic applications in supporting the detection of anxiety disorders efficiently.

Related SDGs Targets

Full Author List

Belluano, Poetri Lestari Lokapitasari; Rahma, Reyna Aprilia; Darwis, Herdianti; Manga, Abdul Rachman, Department of Computer Science, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

This study aims to analyze the performance of various ensemble machine learning methods, such as Adaboost, Bagging, and Stacking, in the context of skin cancer classification using the skin cancer MNIST dataset. We also evaluate the impact of handling dataset imbalance on the classification model’s performance by applying imbalanced data methods such as random under sampling (RUS), random over sampling (ROS), synthetic minority over-sampling technique (SMOTE), and synthetic minority over-sampling technique with edited nearest neighbor (SMOTEENN). The research findings indicate that Adaboost is effective in addressing data imbalance, while imbalanced data methods can significantly improve accuracy. However, the selection of imbalanced data methods should be carefully tailored to the dataset characteristics and clinical objectives. In conclusion, addressing data imbalance can enhance skin cancer classification accuracy, with Adaboost being an exception that shows a decrease in accuracy after applying imbalanced data methods.

Related SDGs Targets

Full Author List

Darwis, Herdianti; Rahmasari, Putri Aulia; Irawati, Program Studi Teknik Informatika, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

Sistem pakar yakni sebuah sistem yang diciptakan berdasarkan keahlian seorang pakar pada bidang terkhusus ke dalam sebuah program komputer. Penelitian ini membahas tentang Sistem Pakar Mendiagnosis Penyakit Gangguan Mental Dengan Metode Certainty Factor. Gangguan mental yakni sebuah keadaan kesehatan yang memengaruhi perasaan, pemikiran, perilaku, serta suasana hati atau gabungan diantaranya. Metode certainty factor dipakai sebagai nilai guna melakukan pengukuran taraf keyakinan penyakit gangguan mental. Penelitian ini bertujuan untuk menghasilkan aplikasi yang bisa memberi bantuan masyarakat dalam melakukan diagnosa dini pada gejala awal penyakit gangguan mental. Pada pengujian akurasi yang dilakukan menghasilkan nilai akurasi pada sistem yaitu sebesar 80% berdasarkan 10 sampel. Aplikasi sistem pakar melakukan diagnosis penyakit gangguan mental telah berhasil diimplementasikan ke dalam sistem memakai metode certainty factor guna mengambil kesimpulan berdasarkan pengetahuan pakar.

Related SDGs Targets

Full Author List

Asfar, A Muh Fitrah; Hasnawi, Mardiyah; Darwis, Herdianti, Informatics engineering, Computer science, Universitas Muslim Indonesia, Indonesia

Abstract

One of the most common causes of death worldwide is skin cancer and its incidence is increasing. To achieve optimal treatment and improve clinical outcomes for patients, precision skin cancer detection and classification approaches are required, which can be achieved through the application of feature extraction and machine learning algorithms. The development of such algorithms to identify important patterns from skin cancer image datasets enables early detection and more accurate classification and more effective treatment. Although previous studies have tried to detect skin cancer using feature extraction techniques such as HFF, HOG, and GLCM, some weaknesses still need to be improved. This research aims to combine various feature extraction methods such as Gray Level Co-occurrence Matrix, Histogram Oriented Gradients, and Local Binary Patterns and machine learning algorithms such as Support Vector Machine, Random Forest, and Gaussian Naïve Bayes in the classification process between Melanoma and Nevus skin cancers. In this research, the number of datasets used is 17,397 derived from the ISIC Dataset. The results showed that the Histogram Oriented Gradients method with Support Vector Machine algorithm achieved the highest accuracy of 92%. The combination of Gray Level Cooccurrence Matrix and Local Binary Patterns with Random Forest algorithm also achieved an accuracy of 92%, the combination of Gray Level Co-occurrence Matrix, Histogram Oriented Gradients, and Local Binary Patterns with Random Forest algorithm also resulted in an accuracy of 92%. These findings suggest that the combination of various feature extraction methods and machine learning algorithms can improve accuracy in skin cancer classification, which in turn can contribute to early detection and more effective treatment.

Related SDGs Targets

Full Author List

Salim, Yulita; Azis, Huzain; Darwis, Herdianti; Purnawansyah,Departement of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia; Kurubacak, Gulsun; Anggreani, Desi;

Abstract

The scope of this book focuses on how information technology may assist in achieving goals and in providing solutions to problems such as a pandemic. Research on the Internet and on technology has been done, and the findings have applications in various sectors that rely on interdisciplinary knowledge. This book explores and describes state-of-the-art research conducted during the COVID-19 pandemic. Topics covered include the IT viewpoint and the rules governing digital transformation throughout the pandemic. The Digital Revolution sped up by a decade during COVID-19, which impacted both the user experience and that of software developers. As a component of the digital transformation process, this book explores the experiences of both the user and developer when attempting to change and adapt while utilizing an information technology program. This book includes five topics: (1) multidisciplinary artificial intelligence, (2) Smart City and Internet of Things applications, (3) game technology and multimedia applications, (4) data science and business intelligence, and (5) IT hospitality and information systems. Each topic is covered in several book chapters with some application in several countries, especially developing countries. The chapters provide insight from contributors with different perspectives and several diverse fields who present new ideas and approaches to solving problems associated with the worldwide pandemic.

Related SDGs Targets

Full Author List

Jabir, Sitti Rahmah; Purnawansyah; Darwis Herdianti; Lahuddin Harlinda; Faradibah, Amaliah; Gaffar, Andi Widya Mufila, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

Nowadays, Indonesia's tourism sector faced challenges in light of the global recession threat. These challenges encompassed high airline ticket prices and inflation, which in turn influenced consumer spending patterns. To tackle these difficulties, the Ministry of Tourism had taken steps to allow foreign investments in the potential tourism object to invest. The involvement of foreign investors had contributed to substantial growth and advancement within Indonesia's tourism industry, thereby presenting numerous opportunities for prospective investors. Indonesia has set a target of attracting more than 7 million foreign tourists by the year 2023, which has increased double from previous year. Based on the literature, the researcher's objective is to analyze the potential of public tourism sites, categorizing them as viable prospects for potential investors. The data had been obtained from Kaggle which the target variable was the rating from 1 to 5. The initial classification attempt, which utilized these five categories, proved unsatisfactory, prompting the application of unsupervised learning techniques to reduce the number of target variable categories. Through the utilization of k-means clustering, the final classification resulted in two overarching categories: “good” and “bad” ratings. Subsequent analysis revealed that Naïve Bayes emerged as the most effective algorithm for this classification task, albeit with no significant difference in results when compared to support vector machines. In conclusion, future research endeavors might consider exploring alternative unsupervised learning methods or conducting more comprehensive feature selection processes before implementing the classification.

Related SDGs Targets

Full Author List

Usman, Fifi Febrianti; Purnawansyah; Darwis, Herdianti; Alwi, Erick Irawadi, Program Studi Teknik Informatika, Fakultas Ilmu Komputer, Universitas Muslim Indonesia, Makassar

Abstract

Potensi kerugian hasil panen diakibatkan serangan penyakit tanaman bawang merah merupakan pemicu utama yang dapat menurunkan produktivitas pertanian.Serangan hama penyakit dapatdiminimalisir dan diatasi dengan cepat apabila petani mampu mengklasifikasikan jenis penyakit yang menyerang tanaman berdasarkan ciri dan gejala yang muncul. Penelitian ini bertujuan untuk klasifikasi penyakit tanaman bawang merah yakni bercak ungu dan moler denganjumlah 320 dataset menggunakan ekstraksi fitur warna Hue Saturation Value dengan metode K−Nearest Neighbor(Euclidean Distance) dan Convolutional Neural Network. Berdasarkan hasil penelitian diperoleh nilai akurasi, f1-score yaitu 94% dan presisi, recal yaitu 97%, 91% pada penyakit bercak ungu sedangkan pada penyakit moler bernilai 94% pada akurasi, presisi, recall, dan f1-score dalam klasifikasi HSV dan KNN. Klasifikasi dengan menggunakan HSV dan CNN menghasilkannilai yang tinggi pada akurasi, precision, recall, dan f1-score dengan nilai 100% di kedua penyakit daun tanaman bawang merah bercak ungu serta moler.Klasifikasi dengan menggunakan deep learning CNNmemperoleh nilai akurasi, presisi, recall dan f1-score yangsangat baik yaitu 100%. Denganuraian ini, klasifikasi penyakit tanaman bawang merah menggunakan HSV dan CNN, dan deep learning CNN dinyatakan mampu mengklasifikasi penyakit tanaman bawang merah yakni bercak ungu dan moler secara efektif dan akurat.

Related SDGs Targets

Full Author List

Febriana, Fina; Salim, Yulita; Darwis, Herdianti; Universitas Muslim Indonesia, Indonesia

Abstract

Kemacetan di Kota Makassar sudah menjadi hal yang lumrah. Hampir setiap ruas jalan yang ada di pusat Kota Makassar mengalami kemacetan akibat pertumbuhan penduduk yang meningkat dari tahun ketahun yang menyebabkan ketidak seimbangan antara jumlah moda transportasi yang ada di jalan raya dengan kapasitas jalan yang tersedia. Dampak yang akan ditimbulkan seperti kemacetan, meningkatnya polusi udara, pelanggaran lalu lintas, dan kecelakaan lalu lintas. Oleh karena itu tujuan dari penelitian ini adalah untuk memprediksi kepadatan lalu lintas di Kota Makassar menggunakan metode Volume Capacity Ratio (VCR). Sumber pengumpulan data yang dilakukan yaitu berupa data primer dan sekunder dengan menggunakan metode analisis data, analisis sistem, dan analisis pengujian. Penelitian ini menghasilkan sebuah sistem yang dapat membantu pihakpihak yang membutuhkan dan dapat mengefisienkan waktu dalam mengatasi kepadatan lalu lintas.

Related SDGs Targets

Full Author List

Purnawansyah, Purnawansyah; Darwis, Herdianti; Universitas Muslim Indonesia, Indonesia; Wibawa, Aji; Yulinda, Nurul; Universitas Negeri Malang, Indonesia

Abstract

AFDAL adalah sistem manajemen pembelajaran yang diterapkan di SMAIT Al-Fityan School Gowa Sulawesi Selatan selama pandemic COVID-19. Penting bagi pihak SMAIT Al-Fityan School Gowa untuk mengetahui bagaimana kepuasan pengguna terkait LMS AFDAL. Penelitian ini menggunakan model End User Computing Satisfaction (EUCS) untuk mengetahui kepuasan pengguna LMS AFDAL dengan mempertimbangkan lima variabel utama yaitu content, accuracy, format, ease of use dan timeliness. Kuisioner dalam bentuk skala likert disebarkan kepada 162 responden yaitu 142 peserta didik dan 20 guru SMAIT Al-Fityan School yang diambil secara acak dan bertingkat. Analisis deskriptif dilakukan sebagai tahap awal dan dilanjutkan dengan uji validitas dan realibilitas yang menyatakan bahwa data valid dan reliabel untuk analisis selanjutnya. Berdasarkan hasil dari uji F diperoleh bahwa variabel input berpengaruh signifikan secara simultan dengan memberikan sumbangsih 98 persen terhadap kepuasan pengguna. Adapun dari uji T diperoleh hasil bahwa setiap variabel memberikan pengaruh signifikan secara parsial terhadap kepuasan pengguna dengan urutan variabel yang paling signifikan yaitu content, timeliness, ease of use, format, dan accuracy yang secara deskriptif juga diperkuat bahwa 76 persen pengguna merasa puas dari segi content dan tingkat ketidakpuasan tertinggi terdapat pada variabel format. Dengan demikian, berdasarkan EUCS diperoleh bahwa peserta didik dan guru sebagai pengguna merasa puas dengan sistem dan pengembangan LMS AFDAL direkomendasikan pada sisi format LMS, yaitu fitur, warna, dan desain.

Related SDGs Targets

Full Author List

Darwis, Herdianti; Umar, Fitriyani; Universitas Muslim Indonesia, Indonesiar

Abstract

Complete data is generally required in data analysis especially in time-series-related study. However, incomplete data due to equipment malfunction, human error, disaster, or other unknown reason is practically discovered. It is required to perform missing data prediction before forecasting the future values. Recommendation system is a system that predicts the "rating" or "preference" of a user over an item. Instead of dealing to a function of time series, the weekly precipitation data of Makassar City is placed into a matrix form consisting of "years" in row as the users and "weeks of the year" in column as the items. This method is also known as matrix decomposition. Accuracy of prediction by root mean square error (RMSE) and mean absolute error (MAE) have been performed to compare the predicted result by using the matrix decomposition to the observed values. In this study, matrix decomposition is discovered as a reliable method in dealing with the missing values of historical observation and forecasting the future values simultaneously

Related SDGs Targets

Full Author List

Ramadhan, Aslan Poetra; Herman, Herman; Darwis, Herdianti, Fakultas Ilmu Komputer Universitas Muslim Indonesia Makassar, Indonesia

Abstract

Computer Based Test (CBT) diera kemajuan teknologi informasi saat ini mulai menjadi pilihan baru terbarukan dalam hal uji kompetensi dalam berbagai instansi. Tidak hanya peruntukannya di dunia pendidikan. Computer Based Test (CBT) mulai ramai diimplementasikan diluar dunia pendidikan sebab efisiensi, efektifitas hingga kecepatan yang ditawarkan Computer Based Test (CBT) menjadi pertimbangan untuk diterapkan. Pada aspek lingkungan, dengan penggunaan CBT, peran Paper Based Test dalam uji kompetensi dapat menekan penggunaan kertas didalamnya. Hanya saja sistem eksaminasi CBT pada umumnya menggunakan sistem pilihan ganda pada proses uji kompetensi. Penelitian ini bertujuan untuk menerapkan Metode Non-Negative Matrix Factorization (NMF) pada sistem CBT dengan model uji kompetensi Essay dimana terdapat jawaban yang variatif dari model ujian ini. Penggunaan Metode Non-Negative Matrix Factorization (NMF) untuk mengolah string jawaban yang dikirimkan untuk dilakukan penilaian secara otomatis, dengan Metode NMF ini dilakukan Analisa hubungan antara sebuah frase/kalimat dengan sekumpulan string yang kemudian dilakukan pembobotan terhadap respon yang dikirimkan untuk diberi penilaian dengan bantuan Generic Relevance of Sentence (GRS).

Related SDGs Targets

Full Author List

Purnawansyah; Herdianti; Alwi, Erick Irawadi, Universitas Muslim Indonesia

Abstract

The potential for yield loss due to shallot plant disease is the main trigger that can reduce agricultural productivity. Pest and disease attacks can be minimized and overcome quickly if farmers are able to classify the types of diseases that attack plants based on the characteristics and symptoms that appear. This study aims to classify shallot plant diseases, namely purple spotting and moles with a total of 320 datasets using Hue Saturation Value color feature extraction using the K-Nearest Neighbor (Euclidean Distance) and Convolutional Neural Network methods. Based on the results of the study, the accuracy, f1-score was 94% and precision, recal was 97%, 91% in purple spot disease while in moler disease it was 94% in accuracy, precision, recall, and f1-score in HSV and KNN classifications. Classification using HSV and CNN yielded high scores in accuracy, precision, recall, and f1-score with a value of 100% in both purple spot and moler shallot leaf diseases. Classification using deep learning CNN obtains very good accuracy, precision, recall and f1-score, namely 100%. With this description, the classification of shallot plant diseases using HSV and CNN, and CNN deep learning are stated to be able to classify shallot plant diseases, namely purple spotting and moles effectively and accurately

Related SDGs Targets

Full Author List

Purnawansyah; Julisa; Darwis, Herdianti; Astuti, Wistiani; Umar, Fitriyani; Sugiarti, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

Mangrove ecosystems play an important role in maintaining the balance of the coastal environment, both from ecological and economic aspects. To support the conservation of this ecosystem, accurate information on the classification of mangrove species is needed. The manual identification process that is still widely used has limitations in efficiency and scale. Therefore, this research proposes a deep learning-based mangrove leaf image classification approach using VGG16 architecture as a feature extractor, which is then classified using two machine learning algorithms, namely Support Vector Machine (SVM) and Random Forest. The dataset used consists of leaf images of three mangrove species, namely Avicennia alba, Rhizophora apiculata, and Sonneratia alba, which were collected directly. To improve the generalization performance of the model, various image augmentation scenarios were performed, including zoom, flip, rotation, and a combination of the three. Performance evaluation was conducted using accuracy, precision, recall, and F1-score metrics using a k-fold cross-validation scheme. The results show that the combination of VGG16 + SVM with zoom augmentation provides the highest accuracy of

Related SDGs Targets

Full Author List

Lahuddin, Harlinda; Muliawan, Muh. Raihan Alif; Darwis; Herdianti; Jabir, Sitti Rahmah; Adawiyah, Rabiatul, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia; Takemoto, Kazuhiro, Department of Bioscience and Bioinformatics, Kyushu Institute of Technology Fukuoka, Japan

Abstract

Cryptocurrency markets exhibit significant volatility and nonlinearity, creating difficulties for precise price prediction. This study assesses and contrasts six deep learning models LSTM, CNN, Transformer, TCN, CNN-LSTM, and TCN-LSTM for forecasting the closing prices of four popular cryptocurrencies: Bitcoin (BTC), Ethereum (ETH), Binance Coin (BNB), and Litecoin (LTC). Each model utilizes historical OHLCV data of 5 years, processed through a uniform preprocessing pipeline that encompasses normalization, sliding window segmentation, and an 80: 20 train-test division. Experimental findings indicate that the hybrid TCN-LSTM model exceeds the performance of all other models based on evaluation metrics including MAE, RMSE, MAPE, and \boldsymbol{R}^{2}, showing its capacity to grasp both short- and long-term time patterns. This study additionally highlights the effectiveness of parallel hybrid architectures, particularly the TCN-LSTM model. These findings contribute to the expanding body of research on deep learning applications in financial forecasting and offer practical guidance for developing robust cryptocurrency prediction models.

Related SDGs Targets

Full Author List

Kurnia, Arif; Harlinda; Darwis, Herdianti, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

Gojek adalah aplikasi layanan on-demand yang telah menjadi salah satu platform terbesar di Asia Tenggara dengan jutaan pengguna aktif dan 5 juta ulasan di App Store. Ulasan ini menjadi sumber informasi penting untuk mengevaluasi dan meningkatkan layanan. Penelitian ini bertujuan untuk menganalisis sentimen ulasan pengguna Gojek dengan mengelompokkan ulasan menjadi lima kelas sentimen: "Sangat Puas", "Puas", "Cukup", "Buruk", dan "Sangat Buruk". Metode yang digunakan meliputi K-Nearest Neighbors (KNN), Naive Bayes, dan Support Vector Machine (SVM). Setelah melakukan text preprocessing, ketiga metode tersebut dievaluasi berdasarkan accuracy, precision, recall, dan F1-score. Hasil penelitian menunjukkan bahwa model SVM dengan kernel Linear mencapai akurasi tertinggi sebesar 79.00%, diikuti kernel RBF dengan precision tertinggi sebesar 83.85%. Model Naive Bayes menunjukkan performa cukup baik dengan akurasi 78.00%, sementara KNN memiliki akurasi terendah sebesar 69.25%. Berdasarkan hasil ini, SVM, khususnya dengan kernel Linear dan RBF, terbukti menjadi metode paling efektif dalam menganalisis sentimen pengguna Gojek, memberikan wawasan yang lebih akurat untuk perbaikan layanan

Related SDGs Targets

Full Author List

Purnawansyah, Department of Electrical Engineering and Informatics, Universitas Negeri Malang, Jl. Semarang, Malang, 65145, Indonesia; Department of Computer Science, Universitas Muslim Indonesia, Jl. Urip Sumoharjo, Makassar, 90231, Indonesia; Darwis, Herdianti; Azis, Huzain, Department of Computer Science, Universitas Muslim Indonesia, Jl. Urip Sumoharjo, Makassar, 90231, Indonesia; Wibawa, Aji Prasetya; Widiyaningtyas, Triyanna, Department of Electrical Engineering and Informatics, Universitas Negeri Malang, Jl. Semarang, Malang, 65145, Indonesia; Haviluddin, Department of Computer Science, Universitas Mulawarman Jl. Kuaro Gunung Kelua, Samarinda, 75119, Indonesia

Abstract

This study aims to assess the effectiveness of various algorithms in the realms of supervised and semi-supervised learning applied to three multiclass facial image datasets: JAFFE, Georgia tech, and Yale. The datasets were partitioned into proportions of 80:20, 75:25, and 50:50 for supervised learning, while semi-supervised learning was conducted with labelled and unlabeled data ratios of 20:80, 25:75, and 50:50. The evaluated algorithms include convolutional neural networks (CNNs), decision tree, long short-term memory, K-nearest neighbors (K-NNs), multilayer perceptron, and support vector classifier (SVC), each with varying parameters. Experimental outcomes reveal that the performance of models depends on the dataset partitioning strategies and the type of algorithms used. Specifically, linear and polynomial SVC consistently yield favorable results in supervised learning, particularly demonstrating efficacy on the Georgia tech dataset. Conversely, on the JAFFE and Yale dataset, linear SVC and K-NN emerge as optimal choices. The inclusion of semi-supervised learning enhances insights, particularly evident in the Georgia tech dataset, where the combination of labeled and unlabeled data significantly improves accuracy, especially when leveraging linear SVC algorithm. Although there are some instances of sub-optimal performance in certain algorithms like CNN on specific datasets, this research provides comprehensive insights into the effectiveness of various models in contexts of limited-label learning. The implications of these findings are crucial in advancing the development of adaptive and robust facial recognition systems, especially in navigating datasets characterized by diverse variations and complexities.

Related SDGs Targets

Full Author List

Fathrurahman, Fauzy; Indra, Dolly; Hasanuddin, Tasrif; Darwis, Herdianti, Universitas Muslim Indonesia, Makassar, Indonesia; Kazuaki, Tanaka, Kyushu Institute of Technology, Iizuka, Japan

Abstract

Background: Waste mismanagement remains a critical issue in Indonesian campuses, where ineffective segregation and collection practices contribute to environmental pollution. Smart technologies offer opportunities to improve waste handling efficiency and monitoring in university environments. Methods: This study developed a smart waste bin prototype that integrates Internet of Things (IoT) sensors, machine learning–based image classification (MobileNetV2 with TensorFlow Lite), GPS tracking, and LoRa communication. The system was designed to classify three types of waste—plastic bottles, snack packaging, and cans—while enabling fill-level monitoring, automated sorting, and real-time location reporting. Results: Experimental results showed strong classification accuracy for plastic bottles (100%), but lower performance for snack packaging (53–80%) and cans (40–67%), especially in low-light conditions or with darker materials. The overall real-time testing accuracy reached 45.1%. LoRa communication provided long-range connectivity but was affected by electromagnetic interference, while GPS tracking was reliable in open areas but inconsistent indoors. Conclusions: The prototype demonstrates the feasibility of integrating AI and IoT for scalable campus waste management. Despite environmental and hardware limitations, it offers a modular framework that can be refined with improved lighting, EMI shielding, and enhanced datasets. This research contributes a practical model for smart campus initiatives and supports the adoption of sustainable waste management practices in higher education environments.

Related SDGs Targets

Full Author List

Darwis; Herdianti, Syahrir, Firdaus Abrazawaiz; Hayati; Lilis Nur, Universitas Muslim Indonesia, Jl. Urip Sumoharjo Km.5, Makassar, 90231, Indonesia

Abstract

Recommendation systems play a crucial role in helping users navigate the overwhelming volume of information on digital platforms. However, conventional Collaborative Filtering (CF) methods often suffer from data sparsity, leading to reduced prediction accuracy and limited recommendation diversity. To address this challenge, this study proposes a hybrid recommendation model that integrates K-Means clustering based on genre, release year, and rating statistics into the Neural Collaborative Filtering (NCF) framework. Unlike previous works that rely on a single dimension like genre or demographics for clustering, our model uniquely combines multiple content-based features. Furthermore, we explicitly integrate the cluster labels as additional embedding features within the NCF framework, enabling more nuanced and context-aware representation learning. Using the MovieLens Latest-Small dataset, our hybrid model significantly outperforms the baseline NCF across all metrics, achieving a Mean Absolute Error (MAE) of 0.6097, a Root Mean Square Error (RMSE) of 0.7946, and improvements in Precision@10 (0.6065) and Recall@10 (0.7063). These findings highlight the effectiveness of our novel, content-aware clustering approach in deep learning recommenders, resulting in more accurate, diverse, and contextually relevant movie suggestions.

Related SDGs Targets

Full Author List

Sombeng, Andi Ranreng; As’ad, Ihwana; Darwis, Herdianti, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

Perkembangan teknologi informasi yang pesat telah menjadikan laptop sebagai kebutuhan penting, terutama bagi mahasiswa dan profesional di bidang teknologi. Banyaknya pilihan laptop dengan berbagai merek dan spesifikasi di pasaran menyebabkan konsumen kesulitan dalam memilih perangkat yang sesuai kebutuhan dan anggaran. Untuk mengatasi masalah tersebut, penelitian ini mengembangkan Sistem Pendukung Keputusan (SPK) menggunakan metode Simple Additive Weighting (SAW) untuk membantu dalam pemilihan laptop yang optimal. Metode SAW dipilih karena mampu memberikan perhitungan nilai alternatif berdasarkan bobot dan skor dari beberapa kriteria seperti harga, prosesor, layar, vga, memori, dan hardisk. Sistem yang dibangun memungkinkan pengguna untuk menentukan preferensi terhadap setiap kriteria, kemudian sistem akan melakukan normalisasi dan perankingan untuk memberikan rekomendasi laptop terbaik. Hasil pengujian menunjukkan bahwa sistem ini dapat meningkatkan efisiensi dan objektivitas dalam pengambilan keputusan. Dengan demikian, SPK berbasis SAW ini bermanfaat sebagai alat bantu seleksi laptop yang informatif, akurat, dan mudah digunakan

Related SDGs Targets

Full Author List

Azis, Huzain; M, Muh Fatwah Fajriansyah; Darwis, Herdianti; Purnawansyah; Hasanuddin, Tasrif; Sugiarti, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

Automatic Speech Recognition (ASR) systems perform well for high-resource languages but remain unreliable for regional dialects with limited data. The Makassar dialect, spoken by approximately nine million people in South Sulawesi, Indonesia, presents unique phonetic and grammatical challenges, including high-frequency particles such as mi, ji, and ko. This study introduces the first benchmark evaluation of state-of-the-art ASR models on Makassar speech. Four models, Whisper (tiny, base, small) and Wav2Vec2 Large XLSR Indonesian, were tested on 305 spontaneous utterances (∼10 minutes) from 10 native speakers. Results show severe performance degradation: the best model (Wav2Vec2 Indonesian) reached 87.73% of word error rate (12.27% accuracy). Error analysis reveals two dominant failure modes: Dialect Particle Blindness (average detection rate 2.9%) and Systematic Phonetic Mismatch (89 vowel confusions), indicating that current models treat dialectal features as noise. These findings underscore the urgent need for dialect-aware ASR adaptation and dataset development, providing a foundation for inclusive speech technology across Indonesia's linguistic diversity.

Related SDGs Targets

Full Author List

Darwis, Herdianti; Adnan, Adam; Purnawansyah; Manga, Abdul Rachman, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia; Wibawa, Aji Prasetya; Irawati, Department of Electrical and Informatics Engineering, Universitas Negeri Malang, Malang, Indonesia

Abstract

The proliferation of online learning platforms has necessitated automated systems for monitoring students' emotional states, given that variations in facial expressions significantly influence engagement and learning outcomes. This study proposes a spatio-temporal classification framework for recognizing emotion intensity levels in the DAiSEE dataset, utilizing Vision Transformer as a spatial feature extractor alongside various temporal models, including LSTM, BiLSTM, TimeSformer, and their hybrid variants. Embeddings are extracted via ViT, after which temporal dependencies are captured by each classifier, incorporating feature-level oversampling to mitigate severe class imbalance. Experimental findings reveal that, despite ViT's ability to generate robust spatial representations, all temporal models struggle to identify minority classes, resulting in predictions biased toward the majority class as evidenced by low balanced accuracy scores and overlapping clusters in t-SNE visualizations. Among all configurations, the ViT + LSTM model delivered the most reliable performance, attaining 59 % accuracy and a 0.59 weighted F1-score on engagement labels, while remaining competitive with prior methods. In essence, integrating spatial and temporal features enhances classification efficacy, yet its effectiveness is substantially constrained by imbalanced data distributions. These results offer a thorough examination of representational challenges in imbalanced affective datasets, along with recommendations for mitigation techniques, crossdataset assessments, and multimodal integrations.

Related SDGs Targets

Full Author List

Ma’arif, A. Muh. Syafei Emil; Fattah; Farniwati, Darwis; Herdianti, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

Dalam upaya untuk meningkatkan Keamanan jaringan komputer di Laboratorium Fakultas Ilmu Komputer UMI, seperti halnya serangan UDP Flooding dan SYN Flooding paket data yang datang sangat banyak dan menumpuk yang bisa saja terjadi kapan saja maka sangat dibutuhkan analisa. Serangan DOS adalah jenis serangan terhadap sebuah komputer atau server dengan cara menghabiskan sumber daya yang dimiliki sehingga tidak dapat berfungsi secara optimal. Sehingga secara tidak langsung menghalangi pengguna lain untuk memperoleh akses layanan dari komputer atau server tersebut. Penelitian ini melakukan klasifikasi serangan pada data-data yang diuji dengan menggunakan metode klasifikasi SVM (Support Vector Machines). Data yang diklasifikasi dari serangan DoS yaitu UDP Flooding dan SYN Flooding dengan mencatat aktivitas data traffic jaringan menggunakan tools Wireshark, Hasil penelitian ini Klasifikasi serangan dengan metode SVM menghasilkan tingkat akurasi yang sangat tinggi dalam waktu perekaman data selama 5 menit mendapatkan data record sebanyak 10.000 data yang sudah di seleksi untuk masing-masing serangan dengan rata-rata class yang diprediksi semuanya menghasilkan akurasi sebesar 100%

Related SDGs Targets

Full Author List

Darwis, Herdianti; Purnawansyah; Umalekhoa, Alfi Syahrin; Adnan, Adam; Umar, Fitriyani; AR, Muh. Aqil Fajar, Universitas Muslim Indonesia, Jl. Urip Sumoharjo KM 5, 90231, Makassar, Indonesia; Salim, Yulita, Universiti Kuala Lumpur, 1016 Jl. Sultan Ismail, Bandar Wawasan, 50250, Kuala Lumpur, Malaysia; Raja, Roesman Ridwan, Kyushu Institute of Technology, 680-4 Kawazu, Iizuka, 820-8502, Fukuoka, Japan

Abstract

This research investigates the application of clustering techniques to network congestion data at Universitas Muslim Indonesia, employing a hybrid metric approach based on packet loss and delay. The study utilized two algorithms, K-Means and K-Medoid, applied in a semi-supervised scenario to group 255,147 network data points into 3, 4, and 5 clusters, considering 10 principal variables. During the pre-processing phase, data cleansing was conducted to address missing values, followed by normalization to standardize the scale of numerical variables, thereby preparing the data for the clustering process. Model validation was performed using four cluster evaluation methods: Gap Statistic, Davies-Bouldin Index, and Elbow Method. The evaluation results indicate that both algorithms were capable of forming valid and reliable clusters. However, the K-Means algorithm demonstrated superior performance compared to K-Medoid, particularly when utilizing three Quality of Service variables: throughput, packet loss, and delay. In this configuration, K-Means yielded more stable clusters, a clearer separation between clusters, and a more structured visualization. Consequently, K-Means is considered more optimal for classifying network congestion levels and presents an effective approach for network data segmentation

Related SDGs Targets

Full Author List

Ikhsan, Muhammad Dirga Juang; As’ad, Ihwana; Darwis, Herdianti, Program Studi Teknik Informatika, Fakultas Ilmu Komputer, Universitas Muslim Indonesia, Jl. Urip Sumoharjo Km.05, Kota Makassar, Sulawesi Selatan, Indonesia

Abstract

Stunting merupakan masalah kesehatan masyarakat yang memerlukan pemahaman terhadap variasi profil balita terdampak. Penelitian ini bertujuan melakukan stratifikasi profil antropometri balita stunting di wilayah kerja Puskesmas Pamboang menggunakan K-Means dengan validasi Silhouette Score. Penelitian menggunakan pendekatan kuantitatif komputasional terhadap data sekunder antropometri balita stunting tahun 2025. Dari 826 data awal, satu pencilan teknis dikeluarkan sehingga 825 data dianalisis. Fitur klasterisasi meliputi umur dalam bulan, berat badan, dan Height-for-Age Z-Score(HAZ), sedangkan jenis kelamin, tinggi badan, kategori TB/U, desa/kelurahan, dan posyandu digunakan untuk interpretasi. Jumlah klaster optimal ditentukan dengan menguji K = 2 hingga K = 10. Hasil tertinggi diperoleh pada K = 3 dengan Silhouette Score0,4011. Klasterisasi menghasilkan tiga profil, yaitu balita dengan gangguan pertumbuhan paling berat, balita stunting usia lebih tua dengan ukuran tubuh absolut lebih besar, dan balita stunting usia lebih muda dengan keparahan relatif lebih ringan. Sebaran wilayah dianalisis pada 10 desa/kelurahan dengan jumlah balita stunting tertinggi. Temuan ini menunjukkan bahwa K-Meansdapat mengolah data antropometri rutin menjadi informasi terstruktur untuk mendukung pemantauan pertumbuhan berbasis profil anak dan wilayah layanan.

Related SDGs Targets

Full Author List

Rezky, Putri Nur; Indra Dolly; Darwis, Herdianti, Universitas MUslim Indonesia;

Abstract

The Program Makan Bergizi Gratis (MBG) is a flagship program of the Prabowo Subianto administration launched in 2024, triggering diverse public responses on social media. Sentiment analysis using deep learning models offers an effective approach to understanding public opinion at scale. However, selecting the appropriate model for Indonesian social media text remains challenging. This study aims to compare the performance of two pretrained transformer models, RoBERTa Base and IndoBERTweet Base, in conducting automatic sentiment labeling on Indonesian tweets related to the MBG program using a zero-shot labeling approach without human-annotated ground truth. A total of 1,831 tweets were collected from platform X and preprocessed using case folding, normalization, and stopword removal. Both models were applied in parallel to label each tweet with sentiment categories (positive, neutral, negative) along with confidence scores. The comparison was evaluated using agreement rate, Cohen's Kappa, and confidence score analysis. RoBERTa Base exhibits a conservative tendency with 75.20% neutral labels, while IndoBERTweet Base produces a more balanced distribution (68.16% neutral). The comparison shows 77.28% agreement with Cohen's Kappa of 0.490 (Moderate Agreement). RoBERTa Base achieves higher confidence (mean: 0.9559, 83.01% above 0.95) compared to IndoBERTweet Base (mean: 0.9236, 68.65% above 0.95). IndoBERTweet Base is more effective in detecting negative sentiment, identifying nearly twice as many negative tweets (13.54% vs. 7.48%). This study recommends IndoBERTweet Base for exploratory research requiring sensitive sentiment detection and RoBERTa Base for precision-critical applications. An ensemble approach combining both models is recommended for production-critical applications

Related SDGs Targets

Full Author List

Anraeni, SIska; Septiawan, M Fiqry; Resa, Muhammad Raihan; Lahuddin, Herlinda; Darwis, Herdianti, Universitas Muslim Indonesia

Abstract

Precise quantification of pineapple (Ananas comosus) cultivation is vital for strategic agricultural decisionmaking and operational efficiency. Traditional manual enumeration methods are resource-intensive, time-consuming, and susceptible to human error. This research presents an automated detection framework utilizing unmanned aerial vehicle (UAV) photography combined with digital image analysis techniques. A YOLO-formatted dataset of high-resolution photos was gathered in Pujananting, Barru, South Sulawesi, Indonesia. An object detection model built on the YOLOv8 architecture was then trained using these pictures. Standard criteria such as precision, recall, and mean Average Precision at a 0.5 Intersection over Union threshold were used to assess the model. Experimental findings indicate the proposed framework achieves excellent detection performance, with mean Average Precision attaining 0.979 and both precision and recall reaching 0.95. The framework was additionally validated for automated plant enumeration, with outcomes aligning closely with manual field observations. These findings demonstrate that UAV-based imaging integrated with deep learning provides an accurate and efficient methodology for pineapple crop monitoring, supporting intelligent agricultural applications and informed management strategies.

Related SDGs Targets

Full Author List

Shabran, Muhammad; Darwis, Herdianti; Anraeni, Siska, Universitas Muslim Indonesia

Abstract

Kesehatan mental mahasiswa menjadi isu penting di lingkungan perguruan tinggi karena berpengaruh terhadap performa akademik, interaksi sosial, dan kesejahteraan psikologis. Mahasiswa sering menghadapi tekanan akademik dan tuntutan sosial yang dapat memicu stres dengan tingkat keparahan berbeda. Penelitian ini bertujuan untuk mengklasifikasikan tingkat stres mahasiswa Universitas Muslim Indonesia berdasarkan instrumen Depression Anxiety Stress Scale-21 (DASS-21) menggunakan algoritma Decision Tree berbasis Classification and Regression Tree (CART) dan Random Forest. Penelitian ini melibatkan lebih dari 3000 responden mahasiswa yang datanya diperoleh melalui kuesioner DASS-21 serta variabel pendukung seperti jenis kelamin, usia, fakultas, indeks massa tubuh, durasi tidur, olahraga, dan penggunaan media sosial. Tahapan analisis meliputi preprocessing data, feature engineering, encoding variabel, penanganan ketidakseimbangan kelas menggunakan Synthetic Minority Oversampling Technique for Nominal and Continuous data (SMOTENC), serta pembagian data 80% untuk pelatihan dan 20% untuk pengujian. Evaluasi model dilakukan menggunakan confusion matrix, accuracy, balanced accuracy, precision, recall, dan Macro F1-score. Hasil penelitian menunjukkan bahwa Random Forest dengan tuning tanpa SMOTENC memberikan performa terbaik dengan accuracy sebesar 0,9065 dan Macro F1-score sebesar 0,8427, serta lebih stabil dibandingkan Decision Tree dalam klasifikasi multi-kelas tingkat stres mahasiswa

Related SDGs Targets

Full Author List

Fadly, Ahmad; Purnawansyah; Darwis, Herdianti, Universitas Muslim Indonesia

Abstract

Media sosial seperti Twitter (X) menghasilkan data opini publik dalam jumlah besar yang dapat dimanfaatkan untuk analisis sentimen. Namun, perbedaan konteks antar topik atau hashtag sering menyebabkan terjadinya domain shift yang dapat menurunkan performa model klasifikasi. Penelitian ini bertujuan untuk menganalisis performa algoritma Naïve Bayes dan K-Nearest Neighbors (KNN) dalam skenario cross-domain menggunakan dua representasi fitur, yaitu TF-IDF dan FastText. Dataset diperoleh dari hashtag #MakanBergiziGratis sebagai domain sumber dan #Danantara sebagai domain target. Metode yang digunakan meliputi preprocessing teks, ekstraksi fitur, pemodelan, serta evaluasi menggunakan accuracy, precision, recall, dan F1-score. Hasil penelitian menunjukkan bahwa kedua model memiliki performa tinggi pada internal test dengan accuracy sebesar 0.94. Namun, pada pengujian lintas domain, Naïve Bayes dengan TF-IDF menunjukkan performa yang lebih stabil dengan accuracy sebesar 0.75, sedangkan KNN dengan FastText mengalami penurunan signifikan, terutama pada kelas negatif dengan nilai F1-score sebesar 0.00. Temuan ini menunjukkan bahwa pemilihan algoritma dan representasi fitur sangat mempengaruhi kemampuan generalisasi model dalam menghadapi domain shift

Related SDGs Targets

Full Author List

Suradi; Atmajaya, Dedy; Darwis, Herdianti, Universitas Muslim Indonesia

Abstract

Penelitian dan publikasi ilmiah merupakan indikator penting dalam menilai kualitas dan reputasi perguruan tinggi, namun Fakultas Ilmu Komputer Universitas Muslim Indonesia (FIKOM UMI) belum memiliki sistem informasi yang mampu menyajikan analisis tren serta prediksi publikasi dan sitasi secara terintegrasi. Penelitian ini bertujuan mengembangkan sistem informasi penelitian dan publikasi berbasis data analitik serta membandingkan performa metode Autoregressive Integrated Moving Average (ARIMA) dan Long Short-Term Memory (LSTM) dalam peramalan deret waktu. Data yang digunakan merupakan data sekunder periode 2017–2025 sebanyak 360 data, dengan tahapan preprocessing, agregasi data tahunan, normalisasi, pembentukan data sekuens, serta pembagian data latih dan data uji (80:20). Evaluasi model dilakukan menggunakan RMSE, MAE, dan MAPE. Hasil penelitian menunjukkan bahwa LSTM memiliki performa lebih baik dibandingkan ARIMA, dengan nilai RMSE 43,16, MAE 40,22, dan MAPE 74,15% pada prediksi publikasi, lebih rendah dibandingkan ARIMA (RMSE 48,71, MAE 46,50, MAPE 84,32%). Pada prediksi sitasi, LSTM juga lebih unggul dengan RMSE 252,44, MAE 250,23, dan MAPE 529,86%, dibandingkan ARIMA (RMSE 288,31, MAE 286,42, MAPE 600,42%). Dengan demikian, sistem yang dikembangkan mampu menyajikan tren serta prediksi yang dapat mendukung pengambilan keputusan strategis dalam meningkatkan kualitas penelitian dan publikasi

Related SDGs Targets

Full Author List

Purnawansyah; Puspitasari, Rahma; Manga', Abdul Rachman; Darwis, Herdianti; Nurhalimah, Sitti, Universitas Muslim Indonesia

Abstract

This community service program aims to improve administrative efficiency in Biji Nangka Village, which previously used manual processes and was prone to delays, inconsistencies, and the risk of archive loss. This activity implemented a website-based self-service system and provided training to village officials on the use of key features such as digital letter management, automatic numbering, and electronic archive storage. A total of 17 participants participated in the training and all successfully operated the system. Evaluation results showed that the time to create letters was reduced from 10–15 minutes to 3–5 minutes. Furthermore, the results of the pre-test and post-test comparison showed a 9.412% increase in participant understanding, indicating the effectiveness of the training in improving the digital competence of village officials. Overall, this program has had a positive impact on improving the quality of administrative services and supporting the realization of digital-based village governance.

Related SDGs Targets

Full Author List

Rahayu, Fitri; Harlinda; Darwis, Herdianti, Universitas Muslim Indonesia

Abstract

Mental health adalah kesejahteraan emosional, psikologis, dan sosial seseorang yang berkaitan dengan cara seseorang berpikir, merasa, dan berperilaku, serta kemampuannya mengatasi stres, menjaga hubungan yang sehat, dan menghadapi tantangan secara positif. Aktifitas di laboratorium sering kali memerlukan waktu dan energi yang cukup banyak, yang dapat meningkatkan tekanan pada mahasiswa dan bisa berdampak pada kesehatan mental mereka. Penelitian ini bertujuan menganalisis sentimen mahasiswa terhadap mental health dalam kehidupan laboratorium dengan menggunakan pendekatan Naïve Bayes Classifier dan K-Nearest Neighbor. Penelitian ini menggunakan beberapa teknik pelabelan, yaitu pelabelan manual dan pelabelan menggunkan NLTK, serta pelatihan dengan 5-fold cross-validation dan penggunaan unigram tokenizing. Hasil penelitian menunjukkan bahwa pelabelan manual dengan pendekatan Naïve Bayes Classifier sedikit lebih unggul dengan tingkat akurasi sebesar 95.58%, presisi sebesar 95.60%, dan recall sebesar 95.54% dibandingkan dengan pendekan K-Nearest Neighbor yang menghasilkan tingkat akurasi 91.66%, presisi 91.83% dan recall 91.49%. Sementara itu, pelabelan menggunakan NLTK dengan pendekatan Naïve Bayes Classifier menghasilkan akurasi tertinggi sebesar 94.11%, presisi 94.05%, dan recall 94.22% dibandingkan dengan pendekatan K-Nearest Neighbor yang memiliki akurasi 90.68%, presisi 90.88%, dan recall 91.01%. Dari hasil pengujian, dapat disimpulkan bahwa pendekatan Naïve Bayes Classifier memberikan hasil yang lebih baik dengan mengklasifikasikan sentimen mahasiswa terkait mental health di kehidupan laboratoriom, dengan pelabelan manual memberikan performa terbaik

Related SDGs Targets

Full Author List

Salim, Yulita, Department of informatics, Universitas Muslim Indonesia, Jl. Urip Sumoharjo, Makassar, 90133, Indonesia; Hidayat, Rivaldi, Department of information systems, Universitas Muslim Indonesia, Jl. Urip Sumoharjo, Makassar, 90133, Indonesia; Darwis, Herdianti, Department of informatics, Universitas Muslim Indonesia, Jl. Urip Sumoharjo, Makassar, 90133, Indonesia; Karnyoto, Andrea Stevens, Bioinformatics and Data Science Research Center, Bina Nusantara University, Jakarta, 11480, Indonesia, Computer Science Department, BINUS Graduate Program - Master of Computer Science Program, Bina Nusantara University, Jakarta, 11480, Indonesia; Puspitasari, Rahma, Department of information systems, Universitas Muslim Indonesia, Jl. Urip Sumoharjo, Makassar, 90133, Indonesia; Mozar, Fitya Syarifa, Bioinformatics and Data Science Research Center, Bina Nusantara University, Jakarta, 11480, Indonesia; Pardamean, Bens, Bioinformatics and Data Science Research Center, Bina Nusantara University, Jakarta, 11480, Indonesia, Computer Science Department, BINUS Graduate Program - Master of Computer Science Program, Bina Nusantara University, Jakarta, 11480, Indonesia

Abstract

Nowadays, facial recognition technology is becoming very important in various real-world applications, and we can easily find it in surveillance systems on various devices or mobile authentication. However, several sophisticated deep learning models require high computational resources and complex training procedures, such as FaceNet and VGG-19. These conditions limit practical applications in low-resource and energy-constrained environments. Based on five simple layers, we used a lightweight architecture for Convolutional Neural Networks (CNNs) in this study by utilizing the standard SoftMax loss function for frontal face recognition that has strong performance for various imaging devices, including webcams, CCTVs, smartphones, and DSLRs. Unlike models that rely on special loss functions or deep architectures, our proposed approach technically balances computational efficiency and high performance for face recognition tasks. A primary dataset that we have collected has been curated, consisting of 23 identity classes, to study the robustness of the model when subjected to varying lighting scenarios and resolutions. It is shown in our study that our model attains a 99.16% accuracy rate and exhibits competitive performance in precision, recall, and F1-score, outperforming or nearly matching FaceNet and VGG-19. This finding confirms that the application of simplified CNN models with standard loss functions is feasible for face recognition, and this method has good accuracy even in resource-constrained environments.

Related SDGs Targets

Full Author List

Luluang, Muhammad Zaqly; Irawati; Darwis, Herdianti, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

This study analyzes the effect of domain filtering on drug recommendation systems based on association rule mining using the FP-Growth algorithm with Neural Collaborative Filtering (NCF) as a comparison. The dataset used was derived from patient medical records containing attributes such as complaints, diagnoses, and drug therapies, with a total of 1,000 patient transactions. To avoid data leakage, the dataset was randomly divided into 70% training data and 30% test data before the modeling process was carried out. Domain filtering was applied by limiting the rule structure so that complaints and diagnoses acted as antecedents and drugs as consequents. The performance of the recommendation system was evaluated using the Precision@5, Recall@5, and Normalized Discounted Cumulative Gain (NDCG@5) metrics. The results of the experiment show that the FP-Growth approach with domain filtering produces higher Precision@5 and NDCG@5 values than the non-filtering approach. The Wilcoxon Signed-Rank test shows that the difference is statistically significant, while effect size analysis using Cliff's Delta shows a practically meaningful impact. Furthermore, a comparison with Neural Collaborative Filtering shows that the collaborative filtering-based approach is less effective on transactional clinical prescription data with limited historical interactions. These findings indicate that integrating medical domain knowledge into FP-Growth can improve the clinical relevance and quality of drug recommendation rankings

Related SDGs Targets

Full Author List

Malik, Muh Aristsyah; Harlinda; Darwis, Herdianti, Universitas Muslim Indonesia

Abstract

This study develops a drug recommendation system using multilabel classification with the Decision Tree algorithm based on patient complaint and diagnosis data from electronic medical records. The dataset consists of patient visit records from a community health center in Pangkajene and Kepulauan Regency and is transformed using multi-hot encoding. Model performance is evaluated under three dataset scenarios (N=500, N=800, and N=1000) using multilabel metrics, including Micro-F1, Samples-F1, Hamming Loss, Jaccard Similarity, Hit@5, Precision@K, and Recall@K. The best Decision Tree model achieved a Micro-F1 score of 0.292, Samples-F1 of 0.281, and Hit@5 of 0.690 on the N=1000 dataset scenario. Bootstrap validation with 1000 iterations indicates relatively stable performance, with narrow confidence intervals across evaluation metrics. These results show that the multilabel Decision Tree model is capable of capturing relationships between patient complaints, diagnoses, and drug therapies while maintaining an interpretable decision structure

Related SDGs Targets

Full Author List

Hasanuddin, Tasrif; R, Muhammad Alif Maulana; Darwis, Herdianti; Gaffar, A. Widya Mufila; Satra, Ramdan; Fattah, Farniwati, Universitas Muslim Indonesia

Abstract

The threat of Distributed Denial of Service (DDoS) attacks is a serious challenge to network stability and security, especially in centralized Software Defined Networking (SDN) architectures. This study aims to evaluate and compare the performance of three classification algorithms in detecting DDoS attacks, namely Fully Connected Neural Network (FCNN), Extreme Gradient Boosting (XGBoost), and Random Forest. The dataset used is public SDN network traffic data that has gone through a normalization and feature selection process to improve model efficiency and accuracy. The evaluation was conducted on four scenarios of training and test data sharing ratios (90:10,80:20,70:30, and 60:40) using accuracy, precision, recall, and F 1-score metrics. The test results show that Random Forest obtained the highest accuracy of 99.61 %, followed by FCNN at 94.95 %, and XGBoost at 99.01 %. Despite the slightly lower accuracy, XGBoost showed the best computational efficiency with the fastest training time of 0.94 seconds and the fastest prediction time of 0.03 seconds, making it an ideal choice for real-time detection system implementation. These findings contribute to the development of a precise, efficient, and adaptive DDoS detection system. Future research is recommended to explore advanced feature selection methods such as Recursive Feature Elimination (RFE) to improve the overall performance and efficiency of the model.

Related SDGs Targets

Full Author List

Marlang, Muh Fatwah Fajriansyah; Darwis, Herdianti; Azis, Huzain; Manga, Abdul Rachman, Universitas Muslim Indonesia;

Abstract

Speech in the Bugis-Makassar dialect poses a major challenge for Automatic Speech Recognition (ASR) systems, particularly in detecting clitic particles that are essential to utterance meaning but do not exist in standard Indonesian. This study investigates the effect of synthetic speech data augmentation using OpenAI's gpt-4o-mini-tts on ASR performance for this dialect, testing four ratios of synthetic-to-real data (0%, 50%, 100%, and 150%) through fine-tuning of the wav2vec2-large-xlsr-indonesian model. Evaluation was conducted using Word Error Rate (WER), Character Error Rate (CER), and Particle Detection Rate (PDR) as a dedicated metric for clitic particle detection. Results show that augmentation at a 50% ratio produced consistent improvements across all metrics compared to baseline, with WER decreasing by 3.47 percentage points and PDR increasing by 3.62 percentage points. Further increases in

Related SDGs Targets