SISINFO : Jurnal Sistem Informasi dan Informatika

Optimizing the New Student Admission Process Through the Implementation of a Laravel-Based CRM System at STABA

Annisa Nur Afinni — 2026-02-27

The New Student Admission Process (PMB) is an important agenda in higher education institutions that requires structured data management and communication to run effectively and efficiently. The Bakti Asih College of Analysts (STABA) still manages prospective student data and promotional activities manually using Microsoft Excel and Google Drive. This approach risks data errors and makes communication monitoring less effective. This study aims to develop a Customer Relationship Management (CRM) system to assist in managing prospective student data, monitoring communication, and supporting targeted follow-up actions. The development method used is Waterfall, which includes needs analysis, design, implementation, verification, and maintenance. The CRM system provides features for managing prospects, WhatsApp integration, chat history, automatic status updates, and promotional team performance reports. Black Box testing shows that all functions work as needed. Overall, this system integrates data management, communication, and promotional performance monitoring into a single structured and controlled platform. The system is also designed to improve data accuracy, responsiveness, and efficiency in managing the overall PMB process.

Analysis of the Usability of the Tinanggea Subdistrict Website Using the Usability Testing Method

Elda Indah Sanda Langi — 2026-02-27

The Tinanggea Subdistrict website provides information and public services for the community at the subdistrict level and surrounding areas. However, the existence of a website must be supported by a good level of usability so that the information provided to users is conveyed properly. This study aimed to analyze the usability level of the Tinanggea Subdistrict website using a usability testing method. The research method used was descriptive with a task-based usability evaluation approach. Testing was conducted by involving eight people who were assigned as users of the website. The usability level was measured using the System Usability Scale (SUS) in the form of a questionnaire consisting of 10 questions. The SUS scores of each user were calculated and averaged to obtain the overall website usability score. The test results show that the Tinanggea Subdistrict website scored 51.1, indicating that it is usable but still not optimal. This study also shows that the website has implemented usability principles, especially in terms of ease of navigation and error prevention. However, the website still needs various improvements, particularly in terms of interaction efficiency, interface design consistency, and interactive service support, so that it can function optimally as a public service medium.

Systematic Literature Review: Machine Learning Algorithm Performance Evaluation of Extract-Transform-Load

Muhammad Faisal Ashshidiq — 2026-02-27

The exponential growth of data in the digital era poses significant challenges for effective data utilization. The Extract, Transform, Load (ETL) process is the foundation for preparing large-scale, unstructured data from various sources (NoSQL databases, log files) for analysis in a data warehouse. However, handling complex data structures such as nested arrays in MongoDB is a major obstacle during the transformation phase. In addition, the purpose of the transformation process is to maintain data quality and integrity. This crucial need requires a robust mechanism for anomaly detection to identify unusual patterns or events that indicate data corruption or system errors. The process of handling system errors requires analyzing nested array data structures using relevant machine learning algorithms for anomaly detection. This literature study is expected to provide valuable insights and identify relevant algorithms in data anomaly detection after the ETL process.

Design and Implementation of a Web-Based Marketing Information System for Oil Palm Midrib Products in the Kayu Raja Village Community

Riski Riski — 2026-02-27

The marketing of oil palm midrib products in Kayu Raja Village remains conventional, limiting market reach and sales efficiency. While numerous studies have examined web-based marketing systems for general MSMEs, limited research has addressed user-centered system design and evaluation for rural communities marketing agricultural by-products such as oil palm midribs. This study aims to design, implement, and evaluate a web-based marketing information system tailored to the characteristics of the Kayu Raja Village community. The research adopts the Software Development Life Cycle (SDLC) using the Waterfall model, encompassing requirements analysis, UML-based system design, implementation, and testing. System evaluation was conducted through black-box testing to assess functional performance and user acceptance testing (UAT) to measure usability and user satisfaction. The findings indicate that the developed system effectively provides product catalog management, online ordering, and structured transaction data processing. User evaluation results demonstrate that the system is functional, user-friendly, and aligned with rural user capabilities. The implementation contributes to expanding market access, improving marketing efficiency, and strengthening the economic competitiveness of the village community.

Comparison of Chi-Square and Information Gain Feature Selection Methods for Support Vector Machine-Based Sentiment Analysis

Vitta Margaret Sinambela — 2026-02-27

Vidio is a local streaming platform that dominates the Indonesian market, but still faces challenges in improving user satisfaction as reflected by its 3.5 rating. To enhance the application, user experience insights are needed, which can be identified through sentiment analysis. This study aims to analyze the sentiment of Vidio application user reviews and compare the performance of the Support Vector Machine model using Chi-Square and Information Gain feature selection. The dataset comprises 4,670 reviews collected from July 01 to November 30, 2024. Model evaluation utilizes Balanced Accuracy metrics optimized through hyperparameter tuning to ensure fair assessment on imbalanced data. The experimental results demonstrate that Chi-Square feature selection yields the optimal performance, achieving a peak Balanced Accuracy of 94.78%. Significantly, this result was attained using a computationally efficient Linear Kernel (). In contrast, the Information Gain method yielded a lower Balanced Accuracy of 94.20% despite utilizing a complex Polynomial Kernel (). These findings conclude that Chi-Square provides a superior trade-off between classification accuracy and model complexity, offering a more robust solution for sentiment analysis.

Activation Function Sensitivity in LSTM-Based Peak Stock Price Forecasting for High-Volatility Financial Time Series

Rizal Rafi Nugraha — 2026-03-12

Stock price prediction remains an intriguing task due to the high volatility and complex temporal dependencies present in financial time-series data. Accurate prediction of the highest stock price is particularly important for investors seeking to identify market peaks and optimize trading strategies. This study investigates the effectiveness of Long Short-Term Memory (LSTM) networks in forecasting DELL’s highest stock price by analyzing the impact of different activation functions. Historical stock price data from 2016 to 2024 were used, and several preprocessing techniques, including data normalization and chronological train-test splitting, were applied. The LSTM models were trained for 100 epochs and evaluated using Root Mean Square Error (RMSE), Mean Squared Error (MSE), and Mean Absolute Error (MAE). The main contribution of this research is a comparative analysis of the sensitivity of LSTM prediction performance to different activation functions, namely ReLU, ELU, Sigmoid, and Tanh, in the context of high-volatility financial time-series data. The experimental results show that the LSTM model using the ReLU activation function achieved the best performance, with an RMSE of 0.557942, MSE of 0.311300, and MAE of 0.338773, outperforming the other activation functions. These findings demonstrate that activation function selection significantly influences LSTM forecasting performance. The results provide practical insights for financial analysts and investors in selecting appropriate deep learning configurations for more reliable stock price prediction.

A Comparative Performance Analysis of Classification Algorithms for Hypertension Diagnosis

Imannudin Akbar — 2026-03-12

Hypertension is a leading cause of cardiovascular diseases, strokes, and kidney failure, with early diagnosis being critical for prevention. Traditional diagnostic methods often face challenges such as human error and inconsistent measurements. While machine learning (ML) has been explored as a potential solution, previous studies have mainly focused on accuracy, often neglecting other important metrics like precision, recall, and F1-score, especially in imbalanced datasets. The primary purpose of this research is to address this gap by comprehensively comparing the performance of four machine learning algorithms - Naive Bayes, Support Vector Machines (SVM), Random Forest (RF), and XGBoost—to provide valuable insights for practical hypertension screening. The dataset consists of 1,985 records with 10 predictor features, including both categorical and continuous variables, and a binary target variable (Has_Hypertension: Yes/No) with a class distribution of 1,032 Yes and 953 No. The data undergoes preprocessing, including categorical encoding and feature scaling for SVM. Models are evaluated using a balanced set of metrics, including accuracy, precision, recall, and F1-score. The results show that RF/XGBoost perform best, with the highest F1 and accuracy, while SVM and Naive Bayes serve as competitive alternatives.

A Gamified Learning Model for Business Intelligence in Technopreneurship Education: Case Study at FTI UNIBI

Acep Hendra — 2026-04-16

This study designs a gamified learning model for the 'Technopreneur Portal' to enhance Business Intelligence (BI) mastery within the Technopreneurship course at Universitas Informatika dan Bisnis Indonesia (UNIBI). While BI is a crucial competency for developing data-driven startups, teaching these complex concepts through traditional, passive e-learning often results in high cognitive load and low student engagement. Furthermore, a significant research gap exists: although gamification is widely used to boost general motivation, its specific integration in teaching technical entrepreneurship and BI remains underexplored. To address this, this study adopts the ADDIE (Analysis, Design, Development, Implementation, and Evaluation) framework to systematically transform the syllabus into interactive, gamified modules. Traditional learning activities are reimagined as "Business Quests," where students earn Experience Points (XP) and specialized "Badges," fostering healthy competition through a class "Leaderboard." The integration of gamification shifts the learning paradigm from passive observation to active, experiential participation, effectively scaffolding the mastery of practical BI skills. This contextually tailored approach significantly boosts students’ motivation, analytical abilities, and overall entrepreneurial readiness. Additionally, the model enables students to build a tangible business portfolio, culminating in a Certificate of Achievement with a QR code for digital, accountable validation of their competency credentials.

Image-Based Cat (Felis Catus) Facial Expression Recognition Using YOLOv8n

Atanasius Manurip — 2026-04-16

Cats (Felis catus) express emotional states through subtle facial cues that are often difficult for owners to interpret accurately. Automated recognition systems can provide objective analysis of these expressions using computer vision techniques. This study proposes an image-based cat facial expression recognition system using the YOLOv8n architecture. A dataset of 794 images was collected and expanded to 3,279 images through data augmentation. Four expression classes were defined based on CatFACS and related frameworks: normal-netral, senang-afiliatif, stres-takut, and marah-agonistik. The model was trained using a 70:20:10 split for training, validation, and testing. Experimental results show an overall mAP50 of 0.728, with the highest performance achieved in the senang-afiliatif class (0.773). However, the marah-agonistik class could not be reliably detected due to severe class imbalance in the dataset, indicating that the current model remains insufficient for recognizing anger expressions in cats. Precision and recall reached 0.939 and 0.94 respectively, indicating reliable detection under confident predictions. The trained model was successfully integrated into a Gradio-based dashboard for real-time expression recognition. These results demonstrate the feasibility of lightweight YOLOv8n for feline facial expression recognition while highlighting that accurate detection of marah-agonistik expressions requires a more diverse and balanced dataset in future research.