Jurnal Kecerdasan Buatan dan Pembelajaran Mesin

Website-Based Convolutional Neural Network for Banana Freshness Detection to Prevent Health Risks

2025-12-25T15:25:55+07:00

Purpose: This study aimed to develop a web-based system to detect the freshness level of Cavendish bananas using image-based classification to help assess fruit quality and reduce losses during distribution.

Methodology/approach: This study used a dataset of 1,030 labeled images of Cavendish bananas categorized into four classes: fresh, ripe, unripe, and rotten. Image preprocessing techniques, such as resizing, normalization, and data augmentation, were applied. A Convolutional Neural Network (CNN) model was trained using Python with machine learning libraries, and the trained model was integrated into a web application built with Streamlit to enable real-time image-based predictions.

Results: The proposed CNN model achieved an accuracy of 99%, with average precision, recall, and F1-score values of 0.99. The web-based system successfully provided real-time freshness predictions through image uploads or camera captures.

Conclusions: This study confirms that the web-based CNN system effectively detects Cavendish banana freshness with high accuracy. The integration into a Streamlit application enables real-time and practical use for quality assessment. The system demonstrates the potential of machine learning to support agricultural quality control, reduce post-harvest losses, and enhance food safety.

Limitations: This study was limited to Cavendish banana images and controlled image conditions; therefore, performance may vary when applied to other banana varieties or different lighting environments.

Contributions: This study contributes to the application of machine learning and computer vision in agricultural quality assessment. The system provides a practical tool for farmers, distributors, and consumers to objectively evaluate banana freshness, reduce economic losses, and support public health.

Air Quality Prediction in DKI Jakarta Using Support Vector Machine: A Comprehensive Classification Approach

2026-06-10T13:06:38+07:00

Purpose: This study aimed to develop a machine learning-based classification model for predicting the Air Pollution Standard Index (ISPU/AQI) categories in DKI Jakarta using the Support Vector Machine (SVM) algorithm. This study specifically addresses the challenge of accurate multiclass air quality classification under real-world urban conditions characterized by high pollution variability and imbalanced class distributions.

Research Methodology: This study employed a quantitative research design using secondary time-series data comprising 1,825 daily observations from the Satu Data Jakarta portal (satudata.jakarta.go.id), covering the period from February to November 2023. Six pollutant parameters, namely PM2.5, PM10, CO, SO?, NO?, and O?, served as predictor features. Data preprocessing included missing value imputation, duplicate removal, label encoding, and min-max normalization. The SVM classifier with a Radial Basis Function (RBF) kernel was implemented using Python’s Scikit-learn library on Google Colaboratory. Hyperparameter optimization was conducted via GridSearchCV with stratified K-fold cross-validation, and the model was evaluated using accuracy, precision, recall, F1-score, and confusion matrix analysis.

Results: The optimized SVM model achieved an overall classification accuracy of 96.1% on the held out test set. The macro-average F1-scores were 86%, 98 %, and 93% for the ‘Good, ’ ‘Moderate, ’ and ‘Unhealthy’ categories, respectively. Class imbalance was identified as a primary challenge, with minority classes (Very Unhealthy, Hazardous) underrepresented in the dataset, contributing to differential model performance across categories.

Conclusions: The SVM-RBF model demonstrated high predictive accuracy for air quality classification in urban tropical environments, confirming its applicability as a foundation for automated real-time air quality monitoring systems. The results establish a replicable methodological framework for similar studies in other Indonesian metropolitan regions.

Limitations: The dataset is geographically restricted to DKI Jakarta and temporally limited to a ten-month period, constraining generalizability. Class imbalance for extreme pollution categories (Very Unhealthy, Hazardous) limits the model reliability for rare but critical events.

Contributions: This study contributes a validated SVM-based classification pipeline for AQI prediction in a tropical megacity context, providing methodological guidance for environmental informatics researchers and urban policy practitioners in developing countries seeking scalable, data-driven air-quality management tools.

SI-MATRE: A No-Code Integrated Application for Used Material Management in a Power Generation Company

2026-06-10T10:27:47+07:00

Purpose: This study develops SI-MATRE, an integrated webbased application for managing used materials at PT PLN Indonesia Power Unit Bisnis Pembangkitan (UBP) Keramasan, a gas and steam power generation unit. The system addresses inefficiencies from manual, paper-based recording, including material accumulation, poor traceability, and delayed disposition decisions.
Methodology: Using the Extreme Programming (XP) agile methodology, development proceeded through planning, design, coding, and testing. Glide Apps, a no-code platform, with Glide Table as backend, enabled rapid prototyping without programming. System artifacts included Context Diagram, DFD, ERD, data dictionary, flowcharts, and UI mockups. Functional modules support material intake, outflow recording, and transaction history with monthly analytics. Seven black-box test scenarios evaluated the system.
Results: All test scenarios passed, confirming system accuracy and completeness. SI-MATRE digitized material movement tracking with OTP-secured authentication, automated code-based data retrieval, and visual analytics. Material accumulation was reduced, and items categorized for repair, auction, or disposal were better organized.
Conclusions: No-code platforms like Glide Apps can effectively deliver industrial inventory solutions, improving transparency, speed, and traceability.
Limitations: Scope is limited to the used material return warehouse, with storage constraints under the Glide free tier, no formal user acceptance testing, and untested peak load performance.
Contributions: This study provides a replicable no-code application framework for industrial material management in energy sector facilities.

Fuzzy Mamdani-Based Book Recommendation System for Academic Library Services: Design, Implementation, and Evaluation

2026-06-10T13:29:29+07:00

Purpose: This study aims to develop and evaluate a Fuzzy Mamdani based book recommendation system for the IIB Darmajaya Library to provide personalized recommendations and improve information retrieval efficiency.
Research Methodology: The system was developed through requirement analysis, design, implementation, testing, and evaluation. A Mamdani Fuzzy Inference System (FIS) was constructed using three input variables: borrowing frequency, book rating, and difficulty level, with recommendation score as the output. Twenty-seven fuzzy rules and triangular membership functions were applied. The system was implemented using PHP, Laravel, MariaDB, and JavaScript, while MATLAB was used for FIS verification.
Results: Functional testing confirmed the successful operation of all system modules, including authentication, recommendation generation, search, filtering, and preference updates. Verification results showed complete consistency between manual calculations and MATLAB outputs, with a sample input producing a recommendation score of 4.3.
Conclusions: The proposed system effectively generated personalized book recommendations and demonstrated the capability of the Mamdani method to manage uncertainty in user preferences.
Limitations: The evaluation was limited to functional testing in a controlled environment. User acceptance testing, scalability assessment, and additional recommendation variables were not examined.
Contributions: This study provides a validated Fuzzy Mamdani recommendation framework and a web based implementation model that can be adapted by academic libraries seeking intelligent recommendation services.

Web-Based Decision Support System for Prosperous Family Classification Using the Analytical Hierarchy Process

2026-06-10T12:01:33+07:00

Purpose: This study aimed to design and implement a web-based Decision Support System (DSS) for classifying prosperous families in Dusun Cibanban, Desa Gerning, Tegineneng District, Pesawaran Regency, Indonesia, using the Analytical Hierarchy Process (AHP) method. The system addresses the inefficiency and subjectivity inherent in the manual family welfare assessment process currently employed by local government administrators.

Research Methodology: The system was developed using the waterfall methodology and the AHP method to assess welfare criteria based on national family welfare standards. It was implemented using PHP (CodeIgniter), JavaScript, and MySQL, with system functionality validated through testing of user and administrator modules.

Results: The system successfully generated AHP-based welfare rankings and passed functional testing. Marsidi achieved the highest priority weight (0.1736), demonstrating the system’s ability to provide objective welfare classifications.

Conclusions: The web-based AHP-DSS system provides a faster, more transparent, and more objective mechanism for prosperous family classification than the manual approach previously employed in Dusun Cibanban. The system successfully digitized the multi-criteria welfare assessment process and produced ranked outputs that are accessible via a standard web browser.

Limitations: The study was conducted in a single sub-village (dusun) with 102 household heads, limiting the generalizability of the findings. The AHP criteria weights were determined through expert elicitation, rather than empirical validation. The system performance under high-concurrency conditions and with larger datasets was not evaluated.

Contributions: This research contributes a validated web-based AHP-DSS implementation for community-level family welfare classification in an Indonesian rural village context, providing a replicable model for local government administrators seeking to modernize social welfare targeting through data-driven decision support.