JKTI Jurnal Keilmuan Teknologi Informasi

Evaluasi Tata Kelola Teknologi Informasi pada Operasional Sistem KRS Menggunakan Framework COBIT 5 Domain DSS04 (Studi Kasus: Universitas X)

Garet Alfirmansyah — 2026-06-13

The poor process of filling out the Study Plan Card (KRS) indicates systemic IT governance problems in many universities in Indonesia. Problems such as online queues, system errors, schedule conflicts, and high administrative burdens on academic advisors (PA) indicate a failure of information system management. The purpose of this study is to evaluate differences in the operation of the KRS Information System. In addition, this study suggests improvements based on the DSS04 (Management Operations) process domain within the COBIT 5 IT governance framework. A qualitative case study with interviews and document analysis at University “X” was used as the research methodology. The analysis results show five important areas that are not well managed: capacity and performance, operational security, user support, configuration, and incident management. The results produce a DSS04 implementation framework that includes the formation of a special operations team, the implementation of a real-time monitoring dashboard, the preparation of standard operating procedures (SOPs), and a tiered training program. According to the evaluation, the implementation of this framework can reduce KRS system downtime by up to 70%, increase user satisfaction (students and PA lecturers) by 40%, and reduce the average incident resolution time from 48 hours to 4 hours. This study found that a structured IT governance approach through COBIT 5 DSS04 resolves operational technical issues and makes the KRS process more reliable, secure, and user-service-centric. Ultimately, this will help the university achieve its academic goals.

Hybrid Modeling to Predict the Impact of Generative AI on Students' Learning Retention

Ardiansyah Ardiansyah — 2026-06-30

The growing adoption of Generative Artificial Intelligence (GenAI) technology in the field of education has sparked global concerns regarding the potential decline in students’ cognitive abilities and the loss of their analytical independence. On the other hand, the majority of previous studies have employed a one-size-fits-all approach that generalizes the impact of artificial intelligence without accounting for the specific behavioral heterogeneity of its users. This gap in the literature serves as the research gap for this study, which proposes Hybrid Machine Learning to predict fluctuations in the Skill Retention Score metric. The K-Means algorithm was implemented to segment the data, predictive modeling using the CatBoost Regressor through SHAP (Semi-Supervised Heterogeneous Adaptive Predictor) explainable AI. The segmentation results confirmed the existence of the following profiles: Cluster 0 (The Heavy Dependent) and Cluster 1 (The Traditional User). Based on these two clusters, the STEM field was found to be the top field in the use of Artificial Intelligence (AI) for the purpose of debugging computational code. Model evaluation revealed that AI adoption behavior variables significantly dictate skill degradation only among extreme users (R² = 0.3131), compared to conventional users (R² = 0.1797). Furthermore, the Shapley value analysis found that AI is proven to be safe as a support assistant or learning assistant if the dependency level remains between 1 and 6; however, if usage exceeds 6, it will affect cognitive retention. In other words, the Shapley value analysis successfully identified the tipping point of the cognitive offloading phenomenon. Nevertheless, a total ban on AI use was found to be ineffective in preserving academic retention scores. Therefore, a transition toward institutional regulations regarding the adoption of artificial intelligence is needed, one that is more adaptive, accountable, and specifically tailored to high-risk demographics, particularly in STEM fields.

A Comparison of EasyOCR and Tesseract Performance in Text Extraction from Digital Images

Abyan Hanif Alfatah — 2026-06-30

Rapid advancements in digital image processing technology have increased the demand for automatic text extraction systems from images, commonly known as OCR (Optical Character Recognition). In this field, there are two widely used tools: and , which are two very popular open-source software packages frequently utilized by developers to meet their needs. However, selecting the most appropriate tool often poses a unique challenge for researchers due to significant differences in the underlying architecture and performance offered by each library. This study aims to conduct an in-depth comparative analysis between Tesseract and EasyOCR, specifically regarding character recognition accuracy and data processing speed under various image conditions. The methodology employed in this study involves collecting a diverse dataset of images, ranging from very clean printed text to images with significant visual noise or distortion. Both software programs will then be tested using the Python programming language to systematically and measurably extract text from the same dataset. Performance evaluation is measured objectively using the Weighted Average Character Error Rate (CER) and Word Error Rate (WER) metrics.

Implementation of Hand Gesture Recognition for Indonesian Sign Language based on MobileNetV2

Radit Widianto — 2026-06-30

Communication is a fundamental human necessity; however, for the deaf community, barriers to interaction
with the general public remain a significant challenge due to limited literacy in sign language. This research aims to
implement a hand gesture recognition system capable of translating the alphabet of the Indonesian Sign Language System
(SIBI) in real-time. The MobileNetV2 architecture was selected as the base model due to its efficiency in processing data on
resource-constrained devices without significantly compromising accuracy. The methodology involves several crucial
stages, beginning with image pre-processing—including resizing and image normalization—to the application of data
augmentation strategies such as rotation, shifting, and brightness adjustment to enhance the model's generalization
capabilities in real-world conditions. The dataset comprises SIBI alphabet classifications from A to Z, collected with high
variability to minimize the risk of overfitting. The results demonstrate that the use of depthwise separable convolutions in
MobileNetV2 allows the system to perform gesture detection with high responsiveness and low computational overhead.
Through hyperparameter optimization, this model is expected to achieve optimal accuracy, providing a practical and
inclusive communication tool for the deaf community within social environments and public services.

Klasifikasi Penyakit Daun Tomat Menggunakan Convolutional Neural Network (CNN) Berbasis EfficientNetB0

Garet Alfirmansyah — 2026-06-30

This research is designed to develop a disease classification system on tomato leaves using the Deep Learning method with a Convolutional Neural Network (CNN) architecture based on the EfficientNetB0 model architecture with a Transfer Learning approach. The dataset used for the research was taken from Kaggle (ashishmotwani/tomato) and consists of more than 20,000 images of tomato leaves divided into 11 classes, namely 10 classes of diseased leaves and 1 class of healthy leaves. The research was conducted using the Google Colab platform with a T4 GPU. The research stages include dataset preprocessing, dataset augmentation, train-validation-test data sharing, building an EfficientNetB0-based CNN model, model training with a fine-tuning mechanism, and performance evaluation using a confusion matrix, precision, recall, and F1-score. The configuration uses an input size of 224 × 224 pixels, a batch size of 32, an initial learning rate of 0.001 and a maximum of 50 epochs. The results show that the best training accuracy of all 50 epochs was 99.63% and the best validation accuracy was 87.27%, achieved in the 49th epoch. The best model was restored from the 49th epoch based on the highest val accuracy value. The classes Powdery Mildew, Target Spot, and Tomato Yellow Leaf Curl Virus obtained the highest recall values. The implementation of EfficientNetB0 has proven effective in classifying tomato leaf diseases, making it suitable for application in mobile and web-based smart agriculture systems.