SISINFO : Jurnal Sistem Informasi dan Informatika

Comparative Analysis of CatBoost, Random Forest, and XGBoost for Predicting Student Academic Performance

Yayu Nur Faidah, Agung Rachmat Raharja, Andri Feriansyah, Ahmad Labudi — Fri, 28 Aug 2026 00:00:00 +0700

Predicting student academic performance has become an important topic in educational data mining because it enables educational institutions to identify students who may require academic support at an early stage. This study compares the performance of three ensemble learning algorithms—CatBoost, Random Forest, and XGBoost—in predicting students' final academic grades using the Student Performance Dataset obtained from Kaggle. The dataset contains 649 student records with demographic, family, behavioral, and academic information. Before model development, the data were preprocessed through categorical feature encoding and feature engineering, resulting in five additional variables: parent_education, previous_grade, total_alcohol, attendance_category, and study_efficiency. The dataset was divided into training and testing sets using an 80:20 ratio. Hyperparameter optimization was performed only for CatBoost using the Optuna framework with 20 optimization trials, while Random Forest and XGBoost were trained using predefined parameter settings. Model performance was assessed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and the Coefficient of Determination (R²). The results indicate that CatBoost achieved the best overall performance, with an RMSE of 1.2888 and an R² of 0.8297, outperforming the other two algorithms. Feature importance analysis also revealed that G2 and previous_grade were the strongest predictors of students' final academic performance. These results suggest that CatBoost is a reliable approach for student performance prediction and can support data-driven academic decision-making.

Development of the Smartcoop Website Using a Rapid Application Development Approach

R. Yadi Rakhman Alamsyah, Arif Bakti Nugraha, Ahmad Dimyati — Fri, 28 Aug 2026 00:00:00 +0700

The SmartCoop Website Management System (WMS) is used as a promotional tool for CV Four Vision Media. An evaluation of the current system revealed limited technology compatibility, weaknesses in access security mechanisms, and suboptimal content management. To address these issues, this study implemented the Rapid Application Development (RAD) method, which allows for an iterative and structured development process through the stages of requirements planning, design, construction, and implementation. The development results demonstrated improved content management flexibility, multilingual content support, and a strengthened authentication system. Functional testing using a black box approach demonstrated that all features performed according to established specifications. These findings demonstrate the effectiveness of the RAD approach in WMS development to improve the quality and reliability of websites used as a promotional tool for companies.

Implementation of an Internet of Things (IoT) Based Digital Residential Gate Security System

Shalina Kalifa Putri, Fahmi Arnes — Fri, 28 Aug 2026 00:00:00 +0700

This study aims to implement an Internet of Things (IoT)-based digital security system for residential gates to improve access management efficiency and real-time monitoring. The system integrates an ESP32 microcontroller, DC gearbox motor, limit switch, Network Time Protocol (NTP), Telegram Bot, Node.js server, MySQL database, and web dashboard to support PIN-based access verification, automatic gate control, real-time status notifications, and digital access activity logging. The main contribution of this study is the practical integration of hardware and software components to support more efficient and manageable residential access security. System testing resulted in 100% functional feasibility and 100% accuracy, while usability testing using the System Usability Scale (SUS) produced an average score of 80, categorized as Grade B (Good). These results indicate that the system performs its main functions according to the specified requirements. However, the evaluation was conducted on a limited prototype scale with a limited number of SUS respondents, so the usability results cannot yet be generalized broadly. The evaluation also did not include quantitative measurements of communication latency and network reliability under short-term and long-term usage scenarios. Future research should involve a larger number of users and evaluate communication latency and network reliability under various operating conditions.

An Explainable CatBoost Framework Optimized with Optuna for Predicting Students' Mathematics Proficiency

Andri Feriansyah, Ahmad Labudi, Yayu Nur Faidah, Agung Rachmat Raharja — Mon, 31 Aug 2026 00:00:00 +0700

Predicting students' mathematics proficiency is an important task in educational data mining because it supports educational assessment and data-informed decision making. This study proposes an explainable machine learning framework by integrating CatBoost, Optuna, and SHAP to classify students into different levels of mathematics proficiency. The study uses the Student Performance in Mathematics dataset from Kaggle, which contains 1,000 student records with demographic, socioeconomic, and academic characteristics. Mathematics scores were grouped into three proficiency categories (Low, Medium, and High), transforming the problem into a multiclass classification task. The dataset was divided into training and testing subsets using an 80:20 stratified split. Hyperparameter optimization was performed on the training data using Optuna with 30 optimization trials and five-fold stratified cross-validation, while the testing dataset was reserved exclusively for the final evaluation. The optimized CatBoost model was compared with XGBoost and LightGBM under identical experimental settings. Performance was evaluated using Accuracy, Precision, Recall, F1-score, ROC-AUC, and Confusion Matrix. The optimized CatBoost model achieved the highest observed performance among the evaluated models, with an accuracy of 80.00%, a weighted F1-score of 80.07%, and a ROC-AUC of 0.9190. SHAP analysis identified reading score, writing score, and gender as the most influential predictors of mathematics proficiency. The results indicate that the proposed framework provides competitive classification performance and model interpretability within the studied dataset.

Decision Support System for Determining School Branch Locations Using the Weighted Product (WP) Method and Multi-Attribute Utility Theory (MAUT)

Nadila Agnestesia Suyadi, Muhammad Dedi Irawan, Fathiya Hasyifah Sibarani — Mon, 31 Aug 2026 00:00:00 +0700

The rapid development of information technology has encouraged educational institutions to adopt data-driven decision-making processes in order to improve accuracy, efficiency, and objectivity in strategic planning. One important decision faced by educational institutions is determining the most suitable location for establishing a new school branch. Sekolah Ikhwanul Muslimin plans to expand its educational services by opening new branches; however, the location selection process is complex because it involves several interrelated criteria, such as land area, land price, population density, number of existing schools, crime level, and public transportation availability. Manual decision-making may lead to subjective judgments and suboptimal location choices, which can affect investment efficiency and long-term institutional sustainability. Therefore, this study aims to develop a web-based Decision Support System (DSS) by integrating the Weighted Product (WP) method and Multi-Attribute Utility Theory (MAUT) as a cross-validation mechanism to verify ranking consistency and reduce potential bias from relying on a single method. The WP method is used to calculate preference values based on weighted criteria, while MAUT is applied to transform criterion values into utility scores so that each alternative can be compared more systematically. The system was developed using the Waterfall model and implemented using PHP and MySQL. Based on the calculation results, alternative A1, located on Jl. Desa Kolam, Percut Sei Tuan, obtained the highest final score of 0.464334 and became the most recommended location among the five alternatives analyzed. The results show that the proposed DSS can assist decision-makers in evaluating school branch location alternatives more objectively, systematically, and measurably.

Usability Evaluation of a Waris Information and Calculation Website Using the System Usability Scale

Mon, 31 Aug 2026 00:00:00 +0700

Limited public understanding of Islamic inheritance (WARIS) principles remains a significant challenge, as many Muslims continue to distribute inheritance equally among heirs without considering the provisions prescribed by Islamic Sharia. This lack of knowledge may lead to inheritance practices that are inconsistent with Islamic legal principles and potentially result in disputes among family members. To address this issue, a web-based WARIS Information and Calculation Website was developed to provide comprehensive educational resources on Islamic inheritance law and to assist users in calculating inheritance distribution accurately according to Sharia regulations. This study aimed to evaluate the usability of the website to determine whether it effectively supports users in accessing information and performing inheritance calculations. A descriptive quantitative approach was employed using the System Usability Scale (SUS) as the evaluation instrument. Data were collected from 89 respondents selected using Slovin's formula from the academic community of Universitas Binaniaga Indonesia. After exploring the website and performing its main functions, participants completed the standard 10-item SUS questionnaire. The findings revealed that individual SUS scores ranged from 67.5 to 95.0, with an average score of 83.46. According to the SUS interpretation scale, the website achieved an Excellent usability rating, indicating that it is intuitive, easy to learn, functionally well integrated, and capable of providing a satisfying user experience. These results demonstrate that the WARIS Information and Calculation Website is suitable for public use and has strong potential to support the dissemination of Islamic inheritance knowledge while facilitating accurate inheritance distribution based on Islamic Sharia principles.

Implementation of Market Basket Analysis on Sales Transactions of a Computer Store Using the ECLAT Algorithm

Fachri Akbar Nasution, Eva Darnila, Hafizh Al Kautsar — Mon, 31 Aug 2026 00:00:00 +0700

Inventory management and product placement practices at Dunia Computer Store are still conducted conventionally, creating the risk of mismatches between inventory availability and customer demand. To address this challenge, a data-driven approach is required to accurately identify customer purchasing patterns. This study implements the Market Basket Analysis (MBA) method using the Equivalence Class Clustering and Bottom-Up Lattice Traversal (ECLAT) algorithm to discover associations among products that are frequently purchased together. The research process includes transaction data collection, data preprocessing, conversion of transaction data into the Vertical TID-list format, frequent itemset mining, and the generation of association rules based on support, confidence, and lift metrics. In addition, a web-based application was developed to facilitate the implementation of the analysis. The dataset consists of 521 sales transactions from Dunia Computer Store covering the period from January 1 to December 30, 2023. The results indicate that McAfee Anti Virus has the highest support value at 23.22%, followed by Mousepad Deskmat at 14.59% and Logitech Mousepad at 11.52%. The generated association rules reveal significant relationships, such as Logitech Mousepad → McAfee Anti Virus with a confidence value of 75.00% and a lift value of 3.23, as well as RAM Corsair DDR4 8 GB → SSD NVME 512 GB Patriot with a confidence value of 70.37% and a lift value of 6.43. These findings demonstrate that the ECLAT algorithm is capable of efficiently identifying customer purchasing patterns. Therefore, the results can be utilized to support cross-selling strategies, product bundling, product placement optimization, and more effective inventory management based on actual transaction data.