JURNAL TEKNOLOGI DAN ILMU KOMPUTER PRIMA (JUTIKOMP)

Implementation of Grid Search Optimization Algorithm and Adaptive Response Rate Exponential Smoothing for Hyperparameter Tuning in Production Activity Determination

Federico Sanjaya, Jesslyn Alvina, Muhammad Amsar Putra, Delima Sitanggang — Wed, 30 Apr 2025 00:00:00 +0000

This research aims to improve the accuracy of production planning at PT Bilah Baja Makmur Abadi by combining the Adaptive Response Rate Exponential Smoothing (ARRES) algorithm and Grid Search optimization. The main problems faced are unpredictable demand fluctuations, dead stock risks, and high operational costs due to imbalances between production and demand. The ARRES algorithm is used for demand forecasting with adaptive exponential weighting, while Grid Search optimizes the alpha and initial year parameters to improve prediction accuracy. This study uses a 5-year sales dataset (2017-2021) with model evaluation using Mean Absolute Percentage Error (MAPE). The results showed that the combination of Grid Search and ARRES optimization algorithms proved effective in helping predict production needs. This can be seen from the significant decrease in the average MAPE value, which is 7.07% using this combination method, compared to 8.18% in the ARRSES method. The lower MAPE value indicates that the Grid Search method is effective in optimizing the ARRSES model parameters. With relatively high prediction accuracy (MAPE < 10%), this method is able to cope with unexpected demand fluctuations.

Application of Support Vector Machine in Measuring Stress Levels Based on EEG Signals

Wed, 30 Apr 2025 00:00:00 +0000

This study aims to classify stress levels based on electroencephalography (EEG) signals using the Support Vector Machine (SVM) algorithm. The data used in this study came from 21 subjects with a total of 379 datasets, which included the main variables of Subject, Electrode Channel (E), Theta, Beta 1, and Beta 2. Preprocessing was done to ensure data quality, including blank data elimination, normalization, and feature engineering. One of the main features developed was the Beta Average, which was obtained by calculating the average between Beta 1 and Beta 2, and stress level classification, which was determined based on the comparison between the Beta Average and Theta. The SVM algorithm was applied to build a stress classification model with an initial stage of manual calculation to understand the basic concepts, followed by the Python programming language implementation. The evaluation results show that the developed model has an accuracy of 92.76%, with the highest precision, recall, and f1-score values reaching 100% and the lowest value of 85%. The confusion matrix analysis showed that the model could classify low stress with 100% accuracy, while it reached 87.8% for high stress. The findings of this study prove that the SVM algorithm effectively classifies EEG signal-based stress levels. This model can be the basis for further development of stress detection methods, especially in mental health and neuroinformatics applications.

Cardiac Abnormality Detection Using Adaptive Neuro-Fuzzy Inference System

Ono Iyan Naibaho — Wed, 30 Apr 2025 00:00:00 +0000

Heart defects are one of the leading causes of death worldwide, making early detection crucial to prevent more serious complications. Electrocardiogram signals are an important diagnostic tool that can be used to detect heart abnormalities in real-time. In this study, an Adaptive Neuro-Fuzzy Inference System artificial intelligence model is used to analyze ECG signal data and detect heart abnormalities early. The ECG signal data used was taken from 30 research subjects, then processed to reduce distracting noise. The combination of artificial neural networks and fuzzy systems aims to overcome the problem of uncertainty in ECG signal data. Thus, this method can be used as a solution that helps in the early diagnosis of heart disorders. The performance evaluation of the proposed Adaptive Neuro-Fuzzy Inference System revealed a perfect True Positive Rate of 1.0 on the Receiver Operating Characteristic (ROC) curve, demonstrating its exceptional ability to correctly identify all instances of cardiac abnormality within the dataset.

Assessment of Plant Growth in Aeroponic Systems Integrated with Smart Farming Compared to Conventional Methods

Muhammad Ihsan, Zulhipni Reno Saputra Elsi — Wed, 30 Apr 2025 00:00:00 +0000

The increasing demand for sustainable agriculture in Indonesia and challenges like land scarcity and inefficient resource use have driven interest in innovative farming technologies. This study investigates the comparative effectiveness of an innovative farming-based aeroponic system versus a conventional soil-based system in cultivating green chili. A quantitative experimental design was employed, using IoT-integrated sensors in the aeroponic setup to monitor and control environmental parameters, while the conventional system relied on manual practices. Key growth indicators, including plant height, number of leaves, and wet and dry weight, were measured over a 30-day day. Statistical analysis revealed that the aeroponic system significantly outperformed the conventional system across all parameters (p < 0.05), with dry weight showing the most substantial improvement. These findings underscore the potential of smart aeroponics in enhancing crop productivity and resource efficiency. However, cost, energy dependency, and scalability considerations must be addressed to enable broader adoption. The study contributes to the growing body of evidence supporting precision agriculture as a viable strategy for sustainable food production.

Design and Development of an Android-Based Application for Hydroponic Introduction and Learning Media

Nur Qodariyah Fitriyah, Ulya Anisatur Rosyidah, Hardian Oktavianto — Mon, 26 May 2025 00:00:00 +0000

Learning media will make it easier for people to learn hydroponic vegetable planting material compared to books because there are visualizations of images and videos so that it will be understood more quickly than just reading. Android is one of the open-source programming languages that allows makers to modify and distribute the results of making applications freely and freely. Developing learning media for hydroponic vegetable planting using Android will make it easier for readers because this application can be used anywhere without carrying more weight than books. In this learning media to be built, the advantages of this application are the visualization of attractive graphics, and simple decision support features related to the implementation of hydroponic farming, which includes capital and available land, as well as suitable plant species. The application has been successfully made, and based on the test results, it can be concluded that it can be appropriately used and produce results according to the design. The suggestions from this research are related to the application's appearance, which can still be developed to be more attractive.

Recent Trends and Innovations in Elementary School Educational Game Development: A Literature Review

Syarifah Atika — Wed, 30 Apr 2025 00:00:00 +0000

Educational games have emerged as interactive learning media that enhance elementary students’ motivation, engagement, and understanding. This study analyzes recent trends in academic game development over the past five years by reviewing 15 peer-reviewed articles published between 2021 and 2025, sourced from Google Scholar. The analysis reveals that 60% of the studies focused on mobile-based games, particularly Android applications developed using Unity and Construct 2, due to their high accessibility and engaging interactive features. Additionally, web-based games such as Wordwall and desktop-based visual novel games developed with TyranoBuilder were found to improve students’ concept mastery by up to 30%, especially in language and mathematics learning. However, key challenges remain, including limited platform compatibility, the absence of adaptive learning features, and weak integration with formal curriculum standards. To enhance their effectiveness, future educational games should prioritize cross-platform accessibility, implement adaptive learning mechanisms, and ensure strong alignment with academic curricullum.

Application of K-Means Clustering Algorithm for Air Quality Pattern Analysis in Jakarta

Wed, 30 Apr 2025 00:00:00 +0000

Air pollution in urban areas, particularly in Jakarta, is a significant issue that impacts public health and environmental quality. This study aims to analyze air quality patterns in Jakarta from 2010 to 2023 using the K-Means Clustering method based on Air Pollution Standard Index (ISPU) data. Data processing stages based on the CRISP-DM methodology are applied to process and analyze data systematically. The stages include business understanding, data understanding, data preparation, modeling, and evaluation. The results showed that the data were divided into three distinct clusters: healthy, unhealthy, and moderate. Cluster 0, which includes stations DKI1 and DKI2, shows better air quality, while cluster 2, which consists of stations DKI3, DKI4, and DKI5, shows higher pollution levels. These findings offer valuable insights for policymakers in developing more effective air pollution control strategies. Thus, the results of this study not only contribute to the understanding of air quality in Jakarta but also emphasize the need for data-driven mitigation actions to improve public health and the environment.

Supply Chain Analysis in the Health Sector Using Gradient Boosting Regression Algorithm

Bayu Angga Wijaya, Nestina Halawa — Wed, 30 Apr 2025 00:00:00 +0000

Supply chain analysis in healthcare is a crucial aspect in ensuring efficient and optimized resource distribution. This study uses the Gradient Boosting Regression algorithm to predict demand in healthcare supply chains to improve the accuracy of stock planning and management trained using supply datasets from hospitals. The model evaluation results show that most of the predictions are close to the actual values, as seen from the points clustered around the reference line. Despite the slight deviations, the Mean Absolute Error (MAE) value of 157.16 indicates that the average prediction error is relatively small compared to the demand scale which ranges from 0 to 14,000. This indicates that the Gradient Boosting Regression model performs reasonably well in estimating supply chain demand in the healthcare sector. Thus, this approach has the potential to be used in more accurate decision-making, in order to improve the efficiency of distribution and availability of health resources