cover
Contact Name
Miftakul Huda
Contact Email
huda.sobari@gmail.com
Phone
-
Journal Mail Official
info@jurnalapik.id
Editorial Address
Hegarmukti, Cikarang Pusat, Bekasi
Location
Kab. bekasi,
Jawa barat
INDONESIA
Journal of Intelligent Systems and Information Technology
Published by Apik Cahaya Ilmu
ISSN : -     EISSN : 30465001     DOI : https://doi.org/10.61971/jisit
Journal of Intelligent Systems and Information Technology (JISIT) focuses on providing scientific articles related to Intelligent Systems and Information Technology, which are developed by publishing articles, research reports and reviews. Journal of Intelligent Systems and Information Technology (JISIT) accepts scientific articles in the field of research: Artificial Intelligence, Data Mining, Text Mining, Web Mining, Machine Learning, Deep Learning, Natural Language Processing (NLP), Social Network Analysis, Expert system, Decision Support System, Computer Network Security related AI, Image processing, Computer Vision, Big Data, and related fields
Articles 33 Documents
Implementation of Machine Learning Algorithm for Credit Scoring Prediction in Islamic Microfinance Siregar, Kiki Hardiansyah; Ruslan, Dede; Faried, Annisa Ilmi; Sembiring, Rahmad
Journal of Intelligent Systems and Information Technology Vol. 2 No. 2 (2025): July
Publisher : Apik Cahaya Ilmu

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61971/jisit.v2i2.156

Abstract

Islamic microfinance institutions face complex challenges in data management and customer behavior prediction in the digital era. This study aims to optimize the Gradient Boosting algorithm with pruning techniques to predict customer collectibility. The analysis was conducted on data from 57 customers with 7 attributes from 2022 to 2024. The research methodology includes four stages: data collection, pre-processing, modeling, and evaluation. Pre-processing involves handling missing data, normalization, encoding, and feature selection. Modeling using XGBoost with and without pruning, followed by evaluation using accuracy, precision, recall, F1-score, confusion matrix, and ROC curve metrics. The results show an increase in model performance with pruning: accuracy increased by 0.70%, precision 0.60%, recall 0.80%, and F1-score 0.70%. This technique is effective in reducing overfitting and increasing model generalization. This research provides significant contribution in developing more accurate credit scoring system for Islamic microfinance institutions, improving credit risk management and customer service in Islamic microfinance sector. The findings help Islamic microfinance institutions optimize credit decision-making process and reduce risk in the digital era.
Design of Forensic Analysis Framework for Single-Page Web Applications Arif Faizal
Journal of Intelligent Systems and Information Technology Vol. 2 No. 2 (2025): July
Publisher : Apik Cahaya Ilmu

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61971/jisit.v2i2.163

Abstract

The Single-Page Application (SPA) architecture based on React, Vue, or Angular has revolutionized browser development with a dynamic user experience. However, its complexity poses challenges for forensic analysis that have not yet been maximally addressed by conventional frameworks. This research designs an open-source forensic testing framework specifically for SPAs, integrating volatile memory acquisition, client-side artifact analysis, and log correlation on APIs. This hybrid method combines live memory forensics with API transaction tracing to reconstruct user activity longitudinally. Evaluation of 15 SPAs showed that the framework successfully identified 92% of XSS and injection traces missed by traditional forensic tools, with a 40% improvement in user session reconstruction. The novelty of the research lies in the SPA-specific modules for virtual DOM extraction and state management, temporal integration using Plaso for synchronizing client-side events and server-side logs, and the adaptation of open-source forensic tools for the SPA environment. In this study, blind spots in forensic activities such as the dynamism of client-side rendering and encrypted API calls can be addressed. Case studies illustrate the effectiveness in uncovering credential hijacking in Vue.js applications. This framework provides a foundation with standards for SPA investigations that critically combine aspects of cloud, client, and API forensics.
Comparative Analysis of Machine Learning Algorithms in Detecting DDoS Attacks on CICIDS2017 Dataset Dika Kurnia Putra; Chandra Ari Pradana; Muhammad Hilal Gilardin; Albert Riyandi
Journal of Intelligent Systems and Information Technology Vol. 2 No. 2 (2025): July
Publisher : Apik Cahaya Ilmu

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61971/jisit.v2i2.182

Abstract

The rapid escalation of Distributed Denial of Service (DDoS) attacks has posed significant threats to global cybersecurity. This research presents a comparative analysis of three supervised machine learning algorithms—K-Nearest Neighbor (KNN), Decision Tree (DT), and Random Forest (RF)—in detecting DDoS attacks using the CICIDS2017 dataset. While many studies focus on broader intrusion detection, this study concentrates specifically on binary classification between benign and DDoS traffic. The CICIDS2017 dataset was chosen for its comprehensive and realistic representation of modern network traffic. The methodology involved preprocessing, training, and evaluating the models in Orange Data Mining using 10-fold cross-validation. Evaluation metrics included Accuracy, Precision, Recall, F1-Score, AUC, and Matthews Correlation Coefficient (MCC). Empirical results show that the Random Forest algorithm outperformed both KNN and Decision Tree, achieving perfect scores across all metrics (1.000). These findings highlight the robustness of ensemble learning in intrusion detection. The results have practical implications for the development of more reliable, efficient, and automated Intrusion Detection Systems (IDS), especially in real-world scenarios prone to volumetric DDoS attacks. Future work should explore multiclass classification and real-time implementation.
Utilizing of Big Data and Machine Learning to Predict the Impact of Sharia Monetary Policy on Economic Stability Kiki Hardiansyah Siregar; Dede Ruslan; Annisa Ilmi Faried; Rahmad Sembiring; Maya Andriani; Ika Swantika
Journal of Intelligent Systems and Information Technology Vol. 3 No. 1 (2026): January
Publisher : Apik Cahaya Ilmu

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61971/jisit.v3i1.243

Abstract

This study aims to investigate the use of big data and machine learning techniques to predict the effects of Sharia monetary policy on economic stability. Motivated by the underexplored nature of Sharia-compliant financial forecasting, the research applies various machine learning models including Random Forest, XGBoost, Support Vector Regression (SVR), Linear Regression, and Long Short-Term Memory (LSTM) networks to long-term stock and monetary trend predictions within the Islamic finance context. Utilizing a dataset of Indonesian Sharia stocks and financial indicators, and employing SHAP analysis for model interpretability, the study evaluates model performance based on metrics like MAPE and R². Results reveal that SVR outperforms other algorithms, providing robust and interpretable predictions that capture the influence of key financial indicators such as moving averages and Sharia-compatible BI Rate transmissions. The findings have practical implications for enhancing financial inclusion and promoting risk-efficient, interest-free profit-sharing models, with national Sharia financial assets projected to exceed Rp10.5 quadrillion by 2026. The study fills gaps in academic literature on Islamic monetary forecasting and sets a foundation for future integration of hybrid AI models for real-time policy evaluation, supporting the digital transformation of Sharia banking aligned with maqasid syariah principles
Design An Effective, Faster Region-Based Convolutional Neural Network with Optimization for The Human Tracking System M Srilatha; N. Srinivasu
Journal of Intelligent Systems and Information Technology Vol. 3 No. 1 (2026): January
Publisher : Apik Cahaya Ilmu

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61971/jisit.v3i1.240

Abstract

Nowadays, real-time Human Tracking System (HTS) is a crucial topic in computer vision and image processing with applications like robotic perception, scene understanding, video surveillance, image compression, medical image analysis, and augmented reality, among many others. In this paper, we design a Faster Region-based Convolutional Neural Network (FR-CNN) with Crow Search Optimization (FR-CNN-CSO) architecture to improve computational complexity and enhance the performance of HTS. The system is implemented in a Python environment with video input. Remove unnecessary data from the gathered datasets during preprocessing. Next, feature extraction is processed using Histograms of Oriented Gradients (HOG). Then update, the extracted features into a designed FR-CNN model for identifying and tracking a person using crow search fitness. The main goal of the developed approach is to attain accurate prediction results and improve the computational complexity by achieving less execution time. Finally, the experimental outcomes show the reliability of the designed system by other conventional techniques in terms of accuracy, precision, recall, F-measure, and execution time.
Blockchain-Enabled Intelligent Platforms: Enhancing Trust and Transparency in Halal Food Kiki Hardiansyah Siregar; Tasriani Tasriani; Dede Ruslan; Devi Megawati; Maya Andriani
Journal of Intelligent Systems and Information Technology Vol. 3 No. 2 (2026): July
Publisher : Apik Cahaya Ilmu

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61971/jisit.v3i2.279

Abstract

Indonesia's halal industry faces significant challenges in bridging the gap between 30 million products requiring mandatory halal certification and only 9.6–9.8 million products officially registered with BPJPH, compounded by 40% consumer skepticism toward halal claims due to a paper-based certification system vulnerable to forgery. Blockchain-enabled intelligent platforms are essential to enhance transparency and trust through real-time digital verification. To design and test a blockchain-based halal verification platform integrated with QR codes and intelligent information systems, enabling end-to-end tracking from producers to consumers, reducing certification fraud, and improving consumer trust and system efficiency. This study employs a quantitative-explanatory R&D approach to develop a three-tier system (frontend web/mobile, Python/SQL backend, blockchain simulation), tested for performance on 10,000–100,000 halal product datasets. Primary data (Likert-scale trust scores, response time, accuracy) and secondary data (BPJPH statistics) were analyzed using descriptive statistics and scalability simulations. Research Findings: The platform achieved 99.7% verification accuracy, 1.2–2.1 second response times, and a 0.3% error rate. Consumer trust scores improved from 3.1 to 4.3 (+39%), with 92% of users rating the system as "very user-friendly." The system scaled effectively to 100,000 datasets with CPU usage below 40%. Blockchain-enabled intelligent platforms effectively enhance consumer trust, certification transparency, and computational efficiency, supporting Indonesia's target of 100% mandatory halal certification by 2030–2035. This platform is ready for integration with BPJPH's ecosystem for national-scale UMKM and halal industries.
MolecuTrace: Development of a Web-Based Molecular Oncology Dashboard for Integrating Sample Metadata, SNV Profiles, and Cancer Stage Classification Muhammad Sobri Maulana; Dwitia Pratiwi; Arditya Prayogi
Journal of Intelligent Systems and Information Technology Vol. 3 No. 2 (2026): July
Publisher : Apik Cahaya Ilmu

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61971/jisit.v3i2.307

Abstract

Cancer molecular profiling generates heterogeneous data, including patient metadata, biological sample information, single-nucleotide variant (SNV) records, and cancer stage labels, which in many hospitals and laboratories remain fragmented across spreadsheets and narrative reports. This study aimed to develop MolecuTrace, a web-based molecular oncology dashboard that integrates sample metadata, SNV profiles, and cancer stage classification. A design science research approach was used to define requirements, model the database, implement the analytics workflow, and evaluate the prototype with a simulated dataset of 120 oncology samples and 253 SNV records. The system comprises a sample registry, a validated manual or tabular SNV import module, an annotation layer, a rule-based molecular review-priority scoring engine, dashboard visualizations, and exportable reports. The dataset had a mean age of 53.9 years, 44.2% metastatic cases, and TP53 as the most frequently altered gene. The dashboard summarized cancer type distribution, sample source composition, top altered genes, metastatic status, and review-priority classes. Variant input is currently limited to manual and tabular entry; support for Variant Call Format (VCF) parsing, external annotation services, and structured clinical-genomic exchange is planned. The score only prioritizes records for expert review and is not a diagnostic, prognostic, or therapy-selection tool. MolecuTrace offers an auditable and clinically readable model for turning molecular oncology data into visual summaries for review, particularly in resource-constrained settings. Clinical validation using real-world data requires prior ethics committee approval.
Security and Performance Analysis of Caesar and Vigenère Ciphers Yandi Anzari; Teuku Djauhari; Tomi kurniawan; Noni Rahmawati; Niko Akbar
Journal of Intelligent Systems and Information Technology Vol. 3 No. 2 (2026): July
Publisher : Apik Cahaya Ilmu

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61971/jisit.v3i2.320

Abstract

While most cryptographic benchmarks prioritize modern algorithms, this study offers a distinct, controlled evaluation of classical monoalphabetic versus polyalphabetic substitution ciphers under systematically scaled data workloads up to 1,000 KB. The objective is to quantify the exact execution throughput and security thresholds of the Caesar and Vigenère ciphers against contemporary automated attacks. The methodology utilizes programmatically built Python modules tested across 100 continuous trials inside an isolated Linux environment, subjected to automated brute-force scripts and statistical Kasiski examinations. The empirical results reveal that both algorithms exhibit linear time complexity, though the Vigenère cipher suffers a 38% to 45% performance overhead due to index alignment. Crucially, while the Vigenère cipher successfully deters primitive brute-force attacks, its structural security collapses catastrophically with a 100% vulnerability rate under statistical cryptanalysis once the file size reaches 50 KB. The practical implication dictates that polyalphabetic substitution provides superficial protection against modern computing power, reinforcing institutional policy to restrict these foundational algorithms exclusively to pedagogical frameworks rather than active information systems.
Application of Data Mining to Predict Student Academic Performance at Universitas Mahakarya Asia Ahmad Budi Trisnawan; Elkin Rilvani
Journal of Intelligent Systems and Information Technology Vol. 3 No. 1 (2026): January
Publisher : Apik Cahaya Ilmu

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61971/jisit.v3i1.153

Abstract

Student academic performance is one of the main indicators of the success of the learning process in higher education. Early prediction of academic performance is important to identify students who have the potential to experience academic difficulties so that interventions can be provided in a timely manner. This study aims to apply data mining to predict the academic performance of students of Mahakarya Asia University based on academic and demographic historical data, by comparing two classification algorithms, namely Random Forest and Support Vector Machine (SVM). The prediction results are in the form of academic performance class labels (variable Y) which consist of two classes, namely “On-Time” and “Delayed”, so that the problem is formulated as a binary classification and not a numerical value prediction. The dataset includes the previous semester's GPA, attendance, core course scores, age, gender, and study program. The research stages include data preprocessing, algorithm implementation using Scikit-learn, 80% training data sharing and 20% test data, and evaluation using 10-fold cross-validation. The evaluation results showed that Random Forest obtained an accuracy of 85.2%, accuracy of 84.1%, recall of 82.7%, and an F1-score of 83.4%, higher than SVM with an accuracy of 78.4%, accuracy of 76.9%, recall of 75.2%, and an F1-score of 76.0%. These findings indicate that Random Forest is superior in handling the complexity of diverse academic data. The model built can be used as a tool for Mahakarya University Asia in identifying at-risk students early and supporting academic decision-making.
Decision Support System for Selecting Large Language Models ChatGPT and Gemini with the SAW Method Kukuh Sulistianto; Sulthan Hilmy Ibrahim; Riccy Widjhaya; David Wongso; Albert Riyandi
Journal of Intelligent Systems and Information Technology Vol. 3 No. 1 (2026): January
Publisher : Apik Cahaya Ilmu

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61971/jisit.v3i1.242

Abstract

The development of artificial intelligence has produced various Large Language Models (LLMs), including ChatGPT and Gemini. Selecting the optimal model is crucial since each LLM offers different strengths and limitations. This study aims to build a decision support system to determine the most suitable LLM using the Simple Additive Weighting (SAW) method. Six criteria were evaluated: answer accuracy, response speed, feature availability, usage cost, data security, and ease of access (UX). Both models were scored on a 1–100 scale through performance observation and evaluation using a controlled simulation dataset across several conversation scenarios. The result shows that ChatGPT obtained a preference value of 0.983, outperforming Gemini with 0.968. Therefore, ChatGPT is recommended as the best LLM for applications requiring balanced performance in response speed, feature availability, data security, and ease of use. This study provides a structured approach for selecting an LLM and opens opportunities for future research in multi-model integration and context-adaptive model switching

Page 3 of 4 | Total Record : 33