Jurnal Ilmu Komputer, Teknologi Dan Informasi

Komparasi Algoritma Fitur Matching SIFT Dan AKAZE Untuk Pencocokan Fitur Wajah Berbasis Citra

Galih Putra Pratama, Husin Fadhil Azizi, Tira Karel Agata, Muhammad Naufal — Sat, 31 Jan 2026 00:00:00 +0000

The problem of matching facial features is an important challenge in biometric systems, especially due to variations in lighting, texture and facial details that affect the stability of keypoint detection. This research aims to compare the performance of the Scale-Invariant Feature Transform (SIFT) and Accelerated-KAZE (AKAZE) algorithms in the facial feature extraction and matching process to determine the trade-off between accuracy and computational efficiency. The dataset used comes from NIST with 393 training images and 341 validation images. Evaluation is carried out using the number of detected keypoints, number of matching keypoints, number of inliers and outliers, feature extraction time, as well as error metrics such as MSE, MAE, RMSE, and R². Experimental results show that SIFT produces better matching performance with a total of 934,763 keypoints detected, an average matching keypoint of 121.14, and the number of inliers of 116.95. In addition, SIFT produces lower MSE, MAE, and RMSE values than AKAZE, indicating better feature matching consistency in facial images. However, AKAZE has higher computational efficiency with an average feature extraction time of 0.1699 seconds, faster than SIFT of 0.2928 seconds. The contribution of this research lies in the comparative analysis of the performance of SIFT and AKAZE in keypoint-based facial feature matching, so that it can be a reference in selecting algorithms according to application needs, both oriented towards accuracy and computational efficiency.

Studi Literatur Review Penerapan Data Mining Untuk Prediksi Penyakit Jantung Menggunakan Naïve Bayes

Anisa Rizki Septia, Hetty Rohayani — Sat, 31 Jan 2026 00:00:00 +0000

Heart disease is one of the leading causes of global death, often difficult to detect early due to non-specific clinical symptoms. To overcome the limitations of manual diagnosis, the application of data mining techniques utilizing the Naïve Bayes algorithm presents an efficient and accurate computational solution. This study aims to analyze and map the effectiveness of Naïve Bayes implementation in predicting heart disease through a Systematic Literature Review (SLR) approach. The contribution of this study is to provide a comprehensive taxonomic guide regarding the influence of data geometry, preprocessing techniques, and the integration of feature selection methods on optimizing the performance of probabilistic models. The results of the literature review indicate that the model accuracy level varies between 58% and 91.80%, with the majority of performance stable in the range of 79%-91% which is deterministically influenced by the quality of data dimensionality reduction. Overall, the Naïve Bayes-based data mining process has proven to have great potential as a clinical decision support system in supporting early medical preventive measures.

Penerapan Metode Convolutional Neural Network pada Identifikasi Wajah Mahasiswa didalam Ruang Perkuliahan

Rahman M. Abdullah, Mohamad Ilyas Abas, Syahrial Syahrial — Sat, 31 Jan 2026 00:00:00 +0000

Manual student attendance systems still present several limitations, including the potential for data manipulation, human error, and low efficiency in large classroom environments. This study aims to implement the Convolutional Neural Network (CNN) method to simultaneously identify students’ faces within a classroom setting. The dataset consisted of 1,740 facial images collected from 58 students using a 2K Full HD webcam under varying capture angles and lighting conditions. The research stages included data collection, image preprocessing, data augmentation, CNN model training, and evaluation using a confusion matrix, accuracy, precision, recall, and F1-score metrics. The developed CNN model, named FACENET V5, was designed using TensorFlow with three convolutional blocks, batch normalization, max pooling, dropout, and a softmax classifier. Experiments were conducted using image sizes of 100×100, 200×200, 300×300, and 400×400 pixels with several dataset split scenarios. The results demonstrated that the 100×100 image size with a 90:10 data split achieved the best performance, obtaining a validation accuracy of 98.28% and a loss value of 0.1127. Furthermore, FACENET V5 was compared with ResNet50V2, MobileNetV2, and VGG16. Comparative results indicated that FACENET V5 provided the most optimal performance in simultaneous student face recognition. This study confirms that CNN can be effectively implemented as an automated face recognition-based attendance system in academic environments.

Komparasi Empat Kernel Support Vector Machine pada Klasifikasi Cyberbullying Twitter Berbahasa Indonesia

Juni Ismail, Randi Sumitro — Sat, 31 Jan 2026 00:00:00 +0000

Cyberbullying on the Twitter social media platform has emerged as a significant social problem in Indonesia, with adverse effects on the mental health and well-being of its victims. Given the enormous volume of daily tweets, automated detection of cyberbullying expressions has become an urgent necessity. This study aims to compare the performance of four kernel functions in the Support Vector Machine (SVM) algorithm namely Linear, Radial Basis Function (RBF), Polynomial, and Sigmoid for cyberbullying classification on Indonesian-language tweets. The dataset used is a publicly available corpus of 13,169 annotated tweets released by Ibrohim and Budi in 2019. The preprocessing pipeline includes case folding, text cleaning, slang normalization using a colloquial dictionary, stopword removal, and stemming based on the Sastrawi library. Text features are extracted using Term Frequency–Inverse Document Frequency (TF-IDF) with a combination of unigrams and bigrams limited to the top 5,000 features. Model training is conducted on a stratified 80:20 split. Experimental results show that the RBF kernel achieves the highest performance with an accuracy of 0.8281 and an F1-score of 0.8269, slightly outperforming the Linear kernel (accuracy 0.8258; F1-score 0.8256). The Sigmoid kernel reaches an accuracy of 0.8204, while the Polynomial kernel records the lowest performance (accuracy 0.7674). The Linear kernel proves to be the most efficient option with the shortest training time (9.19 seconds) without significantly compromising accuracy. These findings can support the development of automated content moderation systems on Indonesian-language platforms.

Peningkatan Akurasi Diagnosis Penyakit Ginjal Kronis melalui Integrasi Algoritma Naive Bayes dan Algoritma Genetika

Eka Pandu Cynthia, Edi Ismanto — Sat, 31 Jan 2026 00:00:00 +0000

Chronic Kidney Disease (CKD) is a significant global health challenge that necessitates early diagnosis to prevent severe organ failure. While machine learning techniques such as Naive Bayes (NB) have been widely implemented for medical classification, their performance is often hindered by redundant and irrelevant features within high-dimensional medical datasets. This study aims to address this limitation by reducing the dimensions of non-contributing medical attributes, thereby minimizing bias and improving classification accuracy. Consequently, this study proposes the integration of Genetic Algorithm (GA) as a feature selection method to optimize the performance of the Naive Bayes (NB) algorithm in diagnosing CKD. The dataset, sourced from the UCI Machine Learning Repository, consists of 400 samples and 24 clinical features. A genetic algorithm was employed to identify the optimal feature subset through a binary evolution mechanism, while NB served both as the classifier and the fitness evaluation function. The results demonstrate that GA successfully reduced the data dimensions by 50%, streamlining the initial 24 features into 12 highly discriminative ones. Evaluation using 10-Fold Cross-Validation revealed a significant increase in accuracy, rising from 92.50% using the standard NB to 98.50% with the integrated GA-NB model. Furthermore, the recall reached 98.40%, indicating the model's high capability in minimizing diagnostic errors for affected patients (false negatives). This research proves that GA-based feature selection effectively enhances diagnostic reliability and model efficiency, presenting substantial potential for implementation in clinical decision support systems for medical professionals.