Komputasi: Jurnal Ilmiah Ilmu Komputer dan Matematika

Prediction and Analysis of Factors Affecting Marketplace Sales Using a Bidirectional LSTM Model for Inventory Estimation

2026-07-12T14:57:44+00:00

Forecasting weekly sales at the product level can help marketplace sellers avoid both excess inventory and stock shortages. The purpose of this study is to investigate the sales forecasting of the Shopee store antianshop by applying the Bidirectional Long Short-Term Memory (BiLSTM) method in conjunction with correlation-based feature selection to predict the inventory. The raw dataset contained 5,426 weekly product records from Shopee Seller Centre covering 22 May 2023 to 31 May 2025; 5,312 records remained after preprocessing. For each product, the data were ordered by week and converted into ten-week input sequences. Pearson correlation showed that Add to Cart (r = 0.5835$) and Enter Cart (r = 0.5579$) were the strongest retained predictors of weekly Units Sold. BiLSTM was then compared with a unidirectional LSTM under the same experimental settings for ten products. LSTM recorded slightly lower average errors, with MAE of 1.7462 units, RMSE of 2.3977 units, and non-zero MAPE of 69.57%, while BiLSTM produced 1.8505 units, 2.4611 units, and 71.18%, respectively. The results indicate that BiLSTM was competitive but did not outperform the simpler LSTM model consistently. We combined the two models in a Streamlit dashboard that shows product forecasts and weekly inventory guidance. Generalizability of the findings should be done cautiously since the analysis used one store, relatively short product histories, and no systematic hyperparameter search.

Multi-Horizon Long Short-Term Memory Forecasting of Digital Product Transactions for Retail Counter

2026-07-27T09:57:08+00:00

In practice, maintaining optimal balance stock at digital product retail counters is challenging due to highly fluctuating and intermittent daily transaction volumes. Consequently, operators frequently rely on subjective judgment, resulting in inaccurate balance allocation decisions. This study addresses this problem by proposing a data-driven forecasting approach centered on a multi-horizon Long Short-Term Memory (LSTM) model to predict daily digital product transactions and support systematic balance planning. Transaction data were collected from MyFastbiller Telegram chatbot logs, covering 739 days with 27,540 valid transaction records transformed into a structured daily time series. Data preparation involved temporal feature engineering, normalization, and 60-day input sequence generation. A two-layer LSTM model was trained for H-1, H-7, and H-30 forecasting horizons and evaluated using RMSE, MAE, and MAPE against standard baseline models (Moving Average, Naive Forecast, Random Forest, and ARIMA). Results demonstrate consistent accuracy, with the LSTM model achieving MAPE values of 15.24% for H-1, 15.14% for H-7, and 16.35% for H-30 all remaining comfortably below the 20% operational tolerance limit and outperforming the baseline methods, particularly on the long-term H-30 horizon. The H-7 horizon achieved the highest precision due to strong weekly seasonality. The primary novelty of this research lies in tailoring multi-horizon LSTM forecasting specifically to real-world intermittent digital retail transactions and deploying the trained model into a web-based operational dashboard for interactive decision support. Overall, the findings confirm that multi-horizon LSTM forecasting provides an accurate and practical foundation for data-driven balance management in small-scale digital retail counters.

Solar Panel Energy Analysis on Remote Control Mobile Robots

2026-07-14T08:37:05+00:00

The ever-increasing energy demand due to technological and industrial developments is driving the search for environmentally friendly energy solutions. Renewable energy, particularly solar energy, offers great potential to replace dependence on dwindling fossil fuels. As a manifestation of innovation in the field of robotics, this Android-based remote-controlled mobile robot is designed to use a 2WP monocrystalline solar panel as the main energy source, with a battery as additional power storage. This robot system integrates an Arduino Uno microcontroller for data processing and an HC-05 Bluetooth module that can control the robot remotely through an Android application. By utilizing solar energy through the solar panel, the robot can reduce dependence on conventional batteries, while increasing energy efficiency. Trials were conducted to evaluate system performance, including measuring the voltage, current, and efficiency of the solar panel. This research aims to contribute to the development of environmentally friendly robotics technology utilizing solar energy. The results of this study indicate that the designed system can work according to plan and has the potential for further development.

Decision Support System for Civil Servant Mutation Using ORESTE

2026-07-07T00:44:26+00:00

This study presents the development of a web-based Decision Support System (DSS) for determining civil servant mutations at the Center for Standardization of Food Crop Instruments (PSITP) Bogor using the Organization, Rangement et Synthèse de Données Relationnelles (ORESTE) method. The existing mutation process was conducted manually using Microsoft Excel, making it susceptible to calculation errors, subjective decision-making, and lengthy processing times. This research aims to develop a system that supports objective, transparent, and efficient employee mutation decisions based on performance and competency assessments. The study utilized data from 40 civil servants collected through interviews, observations, and institutional records. The ORESTE method was applied to rank employees by employing the Besson Rank technique to handle identical criterion values, followed by Distance Score calculations to obtain preference values (Vi) for final ranking. The system was developed using the Waterfall software development model, encompassing requirements analysis, system design, implementation with PHP and MySQL, testing, and maintenance. The resulting application provides role-based access for administrators, employees, and managers, as well as modules for managing employee data, performance and competency evaluations, automated ORESTE calculations, and ranking reports. Functional and validation testing demonstrated that the system operates correctly and produces accurate ranking results. The findings indicate that the proposed DSS effectively improves the objectivity, transparency, and accuracy of civil servant mutation decisions while reducing manual errors and processing time, thereby supporting more reliable and efficient personnel management.

Implementation of Artificial Intelligence in Health Screening Systems for Category-Based Fitness and Nutrition Recommendations

2026-07-16T04:11:04+00:00

The increasing prevalence of non-communicable diseases, such as diabetes, hypertension, and metabolic disorders, highlights the need for accessible health screening services. However, existing health screening systems generally provide examination results without offering automated educational recommendations to support preventive healthcare. This study proposes an Artificial Intelligence (AI)-based health screening information system that automatically generates fitness and nutrition recommendations based on categorized health screening results. The research employed a Research and Development (R&D) approach using the Extreme Programming (XP) software development methodology. The proposed system was developed using the Laravel framework and the Filament administration panel and integrates a Large Language Model (LLM) through the OpenAI API. The proposed architecture combines structured prompt engineering, predefined AI guardrails, SHA-256 prompt hashing, and recommendation caching to improve recommendation consistency and computational efficiency. The system was evaluated using User Acceptance Testing (UAT) and an AI Recommendation Consistency Evaluation. The UAT results showed that all functional requirements were successfully fulfilled. The consistency evaluation demonstrated that repeated processing of identical health screening data produced stable recommendations, achieving an average qualitative consistency score of 90.4% and an average TF–IDF cosine similarity of 0.784. These findings indicate that the proposed architecture is capable of generating reliable and consistent educational recommendations while reducing redundant AI requests through recommendation caching. This study contributes to the development of intelligent health information systems by introducing an efficient AI recommendation architecture that supports digital health screening and preventive healthcare services.

Assessment of Information Security Maturity Using the KAMI Index 5.0 Aligned with ISO/IEC 27001

2026-07-01T07:56:07+00:00

The advancement of digital government initiatives has led to a wider deployment of IT systems for public service delivery. Consequently, public sector agencies must establish robust cyber defenses to safeguard critical information and infrastructure. This research evaluates the information security posture of the Padang City Communication and Informatics Office, focusing on its readiness to sustain digital transformation. The assessment aligns with the ISO/IEC 27001 framework and utilizes the KAMI Index 5.0 as the primary diagnostic tool. Employing a qualitative descriptive methodology, the study collected empirical data through field observations, stakeholder interviews, and comprehensive policy reviews. The diagnostic results revealed a compliance score of 518, demonstrating that the organization has established the baseline requirements of the ISO/IEC 27001 standard. Nonetheless, the overall maturity remains at Level II, showing that while several protective processes are active, they lack formal documentation and optimal coordination. To bridge these gaps, structured action plans are formulated across seven key assessment domains: governance (5 actions), risk mitigation (14 actions), security policy framework (5 actions), asset classification (13 actions), technical infrastructure (6 actions), data privacy (11 actions), and third-party control (6 actions). These actionable steps provide a strategic roadmap to enhance the mature of city's electronic administration security.

Decision Tree-Based Anomaly Traffic Detection for Local Area Network (LAN) Security Using Wireshark and Nmap Data Analysis

2026-06-28T07:31:27+00:00

The development of information technology encourages the use of LAN networks as primary infrastructure in educational environments. The high intensity of network usage at the Faculty of Engineering and Computer Science increases security risks such as unauthorized access, open ports, and anomalous traffic, making comprehensive network security analysis necessary. This study aims to analyze LAN network security at the Faculty of Engineering and Computer Science using MikroTik (RB951Ui-2HnD) and TP-Link (TL-WR841N) devices, and to apply the Decision Tree algorithm for network traffic classification. The methods include router configuration analysis, port scanning using Nmap (Zenmap), packet sniffing analysis with Wireshark, and network traffic classification using the Decision Tree algorithm implemented in RapidMiner Studio. Port scanning results indicate that from 1000 scanned ports, only 5 ports (0.5%) were detected as open. Wireshark packet capture over 5 minutes collected 6,708 packets, revealing the presence of unencrypted HTTP packets and TCP errors. The Decision Tree model achieved an accuracy of 86.05%, precision of 99.94%, and recall of 73.85% in classifying normal and anomalous traffic. This approach effectively provides an overview of LAN security conditions and can serve as a reference for improving network security in educational institutions.

The Landscape Image Classification Using Convolutional Neural Network on Intel Image Classification Datase

2026-07-07T03:50:47+00:00

Image classification is an important area of computer vision and artificial intelligence that enables computers to automatically recognize and categorize visual information. This research aims to develop a Convolutional Neural Network (CNN)-based image classification model for recognizing six categories of natural and urban landscapes using the Intel Image Classification dataset from Kaggle. The preprocessing stage included image resizing, data augmentation, and pixel normalization to improve model generalization and reduce overfitting. The dataset was divided into 80% training data and 20% testing data. The proposed CNN architecture consists of four convolutional layers, max-pooling layers, and three fully connected dense layers with ReLU and Softmax activation functions. The novelty of this study lies in the development of a lightweight CNN architecture that achieves competitive performance without relying on pretrained models or transfer learning approaches, making it suitable for deployment on resource-constrained devices. Experimental results show that the model achieved 85.85% training accuracy and 85.47% testing accuracy. Performance evaluation using precision, recall, F1-score, and confusion matrix indicates balanced classification performance across all classes. Furthermore, the trained model was successfully converted into TensorFlow SavedModel, TensorFlow Lite, and TensorFlow.js formats to support cross-platform deployment. The findings demonstrate that the proposed CNN model is effective, efficient, and suitable for real-world landscape image classification applications.

The Integrating Business Intelligence and Food Supply Chain Analytics to Support National Nutrition Programs (MBG): Evidence from Rice Production Dashboard in Indonesia

2026-07-19T00:35:31+00:00

This study provides the development and implementation of a food supply chain analytics-based Business Intelligence (BI) dashboard for assisting the Indonesian national nutrition programmes, specifically in the monitoring of rice production. Rice is the number one staple food and a key part of nutrition strategies implemented through mass interventions like the National Nutrition Fulfillment Program, therefore, guaranteeing data-driven decision making all along the rice value chain is crucial. The data for this research are secondary data covering rice production, harvested area, rice productivity 2020 – 2024 processed and visualised with the software Tableau Public. The proposed dashboard is a combination of various analytical views: such as temporal trend analysis (line graph), spatial distribution mapping (geographical visualization), comparative regional performance (bar chart) and proportional productivity assessment (donut chart). These visualisations allows stakeholders to discover production patterns and regional inequalities in an interactive and concise format, and possible supply risks. The findings show the efficiency of the BI dashboard in visualizing large amounts of agricultural data into actionable information, which can be used to guide and shape agriculture strategies and policy formulation. The system successfully identifies the major provincial differences in harvested area as well as the differences in production areas and over time, and is important for adopting food production to meet the operational requirements of the nutrition service unit. This research adds to the extensive research work about the parallel between Business Intelligence and public policies and its applications to data-driven agriculture and food supply chain analytics. The study highlights the opportunities of BI-based dashboards to improve transparency, efficiency and responsiveness in national level food systems to foster nutrition program sustainability and scale. This approach can be further enhanced by incorporating real-time data sources and predictive analytics in future studies to further improve the decision support process.

Application of CPM and PERT Methods in Residential Construction Project Scheduling: A Case Study

2026-07-27T10:38:55+00:00

This study aims to analyze and optimize the scheduling of a residential construction project using the CPM, crashing techniques, and the PERT. A quantitative approach is employed by utilizing data obtained from observations and interviews conducted on a housing construction project in Bandar District, Simalungun Regency. CPM analysis is used to identify the critical path and determine the project duration, while crashing techniques are applied to obtain acceleration alternatives by considering additional costs. Furthermore, the PERT method is used to analyze uncertainty in project duration and to calculate the probability of project completion. The results show that the project duration under normal conditions is 59 days, with the critical path identified as A–B–D–E–G–H–J–O–P. After applying crashing, the project duration is reduced to 55 days with an additional cost of IDR 212,500. The PERT analysis indicates that the expected project duration is 44 days, with a completion probability of approximately 99% for a target duration of 49 days. These findings demonstrate that the combination of CPM, crashing, and PERT methods is effective in improving time and cost efficiency, while also providing a more reliable basis for decision-making under conditions of uncertainty.

A Multi-Ontology Approach for Deforestation Issue Analysis in Indonesian Digital Media

2026-07-27T09:36:36+00:00

The increasing volume of online news discussing deforestation has created challenges for automatically identifying and categorizing environmental issues because conventional text classification methods often fail to capture semantic relationships between terms. This study proposes a multi-ontology semantic feature extraction framework that integrates DBpedia, IndoWordNet, and a domain-specific keyword ontology to enrich document representation for Indonesian deforestation issue classification. News articles were preprocessed through case folding, tokenization, stopword removal, and stemming. Semantic scores generated from the three ontology resources were combined to produce semantic feature vectors and initial issue labels, which were subsequently classified using a Support Vector Machine (SVM). To address class imbalance, Synthetic Minority Over-sampling Technique (SMOTE) and class-weight balancing were incorporated during model training. The proposed model was evaluated using 5-fold stratified cross-validation on a dataset of 567 labeled sentences collected from Indonesian online news portals. Experimental results achieved an accuracy of 73.24%, a macro precision of 61.39%, a macro recall of 47.63%, and a macro F1-score of 51.55%. These results indicate that the proposed multi-ontology framework provides richer semantic representations for deforestation issue classification and offers an interpretable approach for supporting large-scale analysis of Indonesian online news.

The Neural Network Based Multiple Linear Regression Model for Financial Profit Prediction

2026-01-06T03:22:30+00:00

Accurate financial forecasting plays a vital role in supporting business decision-making, particularly for organizations that depend on operational efficiency and strategic planning. CV. Surya Cipta Estetica Mandiri still manages its financial records manually, resulting in inaccurate estimations and limited predictive capabilities. This study proposes a profit prediction model using a Neural Network–based Multiple Linear Regression approach implemented through a Single Layer Perceptron (SLP) and integrated into a financial information system.

The dataset consists of 3,127 daily transaction records from January 2021 to April 2025, including date, category, total income, and total expense. The preprocessing phase includes handling missing values, label encoding, RobustScaler-based feature scaling, and an 80:20 train–test split. The model was developed using TensorFlow with a Sequential architecture, trained using 2,501 data points and evaluated using 626 data points.

The model demonstrates excellent predictive performance, achieving R² = 0.9988, MAPE = 1.03%, MAE = 258,080, RMSE = 893,381, and an overall accuracy of 98.97%. These results indicate that the SLP-based Multiple Linear Regression model provides stable and highly accurate profit predictions. Furthermore, the model can be expanded into a broader ERP system by integrating additional business modules such as finance, HR, and inventory management. Future research may incorporate external market variables, adaptive learning mechanisms, and non-linear algorithms such as Random Forest or Gradient Boosting to improve model robustness and predictive generalization.