The rapid advancement of deep learning significantly increases computational demands, making performance optimization essential for model scalability and deployment. While numerous studies optimize neural network architectures, the effect of different programming paradigms on computational efficiency remains insufficiently explored. This study aims to compare Object-Oriented Programming (OOP) and Data-Oriented Programming (DOP) paradigms in TensorFlow-based deep learning workflows, focusing on their performance across four processing phases: build, compile, train, and evaluate, under a controlled experimental environment with repeated iterations and systematic measurements. Both paradigms are implemented using identical Convolutional Neural Network (CNN) architectures trained on the CIFAR-100 image dataset over thirty controlled experimental iterations. A custom profiler integrating the Python System and Process Utilities (psutil) and NVIDIA Management Library (pynvml) monitors real-time system performance, capturing CPU and GPU utilization as well as memory usage. The results reveal that DOP achieves better resource efficiency with lower memory usage (549.98 MB versus 676.25 MB), higher GPU utilization (64.68% versus 61.08%), and faster evaluation execution (1.50 seconds versus 2.59 seconds), while also attaining higher model accuracy (32.38% versus 28.08%). In contrast, OOP benefits from TensorFlow’s Sequential API optimizations, resulting in faster training times but greater CPU and memory consumption. These findings highlight that DOP provides superior runtime efficiency and offers practical benefits for performance-critical deep learning applications.
Digital transformation in the financial industry has encouraged organizations to adopt application integration systems to improve operational efficiency and effectiveness. However, not all information system implementation projects succeed in meeting expectations, particularly in guarantee institutions with complex business processes. This study evaluates the success of an application integration system project in a guarantee company using a combined approach based on the DeLone and McLean (2003) Information System Success Model and Critical Success Factors (CSF). By explicitly incorporating CSF variables as antecedents of system quality, information quality, and service quality, this study extends the conventional application of the DeLone and McLean model by integrating key managerial perspectives into the evaluation of integration success. A quantitative method was employed using survey data collected from 120 respondents of the Penjaminan Application Integration (PAI) system at PT XYZ. The data were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM). The results indicate that CSF variables, including top management support, internal communication, user training, and project risk management, significantly affect system quality, information quality, and service quality. Furthermore, these three quality dimensions within the D&M model notably influence the intention to use and user satisfaction, which ultimately impact the perceived net benefits for the organization. In conclusion, considering both managerial and system quality perspectives provides a more comprehensive understanding of the success factors in application integration projects. These findings can serve as a basis for improving IT project implementation strategies in the guarantee sector and other industries with similar characteristics.
Food commodities are essential in developing countries, such as Indonesia. The government rules the food commodity prices in every province. Yet, somehow, there are some issues in certain provinces. Data science and statistics techniques can help the government control food commodity prices. The proposed method applied to predict the commodity price index in East Java is the STACKEL K-MEANS method. This proposed method is a collaborative framework that utilizes cluster analysis and stacking ensemble learning to predict data. Cluster analysis is performed first, using the distance that suits time series data, which is Dynamic Time Warping. Two clusters are formed from each commodity (Rice, Oil, and Flour). Then, the stacking model consists of a base learner and a meta learner. The base learner models used are Ridge Regression, Random Forest, and Support Vector Regression, while the meta learner is Light Gradient Boosting Method. To optimize the parameter, we used a grid search. Following the evaluation process, we compare the proposed method with auto ARIMA from Python. In training and testing data, The proposed method yields superior results to the ARIMA model across all three error metrics: MAPE, MAE, and RMSE. The following scores for flour commodities are 0.042% compared to 0.328%, 4.715 compared to 37.57, and 6.34 compared to 523.99. For rice commodities, the scores are 0.261% compared to 0.392%, 31.585 compared to 48.142, and 41.92 compared to 56.068 For oil commodities, the scores are 0.185% compared to 0.250%, 33.02 compared to 47.571, and 39.35 compared to 56.060.
Balinese language is a local language that is widely use and spoken by Balinese people including in social media. However, the nuances of these politeness levels are often lost in informal digital communication and there is a significant lack of computational model to automatically classify them, especially for low-resource language like Balinese. The primary objective of this study is to evaluate the performance of the Multinomial Naive Bayes method combined with Term Frequency-Inverse Document Frequency (TF-IDF) feature extraction, Chi-square feature selection, and Synthetic Minority Oversampling Technique (SMOTE) in classifying Balinese language levels. The dataset for this study consists of 1,314 annotated social media posts and comments, primarily sourced from Instagram. The annotation was conducted by a Balinese language expert to categorize text into six levels that represent varying degrees of politeness and formality. These levels are alus singgih (polite, used for respecting others), alus sor (polite, used for self-humbling), alus mider (polite, used for both respecting others and self-humbling), alus madia (an intermediate level of politeness), basa andap (casual, commonly used in everyday life), and basa kasar (impolite, often used during arguments or toward animals). The experimental results showed that the model successfully achieved an accuracy of 96.53% on the training data and 61.45% on the test data. Additionally, hyperparameter tuning revealed that the Multinomial Naive Bayes model with 2,720 selected features and SMOTE oversampling achieved an accuracy of 91.78%, significantly outperforming the baseline model without feature selection and oversampling, which obtained only 64.93% accuracy.
Stunting is a form of chronic nutritional deficiency in toddlers and remains a major public health concern due to its impact on child growth and development. Efforts to reduce its prevalence continue to be strengthened in Indonesia, particularly in Sumatra Province. This study aims to evaluate the accuracy of a logistic regression model and three machine learning models—decision tree, random forest, and support vector machine (SVM)—in classifying stunting prevalence. The response variable is defined as the prevalence of stunting among toddlers, categorized into two classes: exceeding the national target and not exceeding the national target, based on the 2024 national threshold. Although classification models can provide accurate predictions, they often lack interpretability. Therefore, this study applies the SHAP method to the best-performing machine learning model to identify the key factors influencing stunting. The use of Shapley values is justified through the uniqueness theorem, which establishes it as the only attribution method satisfying desirable fairness properties. SHAP values are employed to explain the model by referencing both the trained model and the underlying data. The results show that the random forest model achieves the highest accuracy (90.00%), outperforming the other models. SHAP analysis reveals that Underweight is the most influential predictor contributing to stunting prevalence in Sumatra Province. These findings highlight the relevance of machine learning interpretability in supporting policy decisions for stunting reduction.
Identifying highly potential athletes is a critical yet inherently challenging process that requires comprehensive analysis of diverse factors, including physiological attributes, demographic characteristics, and social influences. This multifaceted process requires meticulous evaluation of extensive datasets to ensure both accuracy and fairness in talent identification protocols. The complexity stems from the interconnected nature of the determinants of athletic performance, where physical capabilities intersect with psychological resilience, social support systems, and environmental factors. In recent years, machine learning (ML) algorithms gain prominence in decision-making processes, offering unprecedented opportunities to uncover subtle patterns and relationships within athlete data that might otherwise remain hidden. This study systematically benchmarks the performance of several state-of-the-art ML classifiers using a novel, self-collected dataset of athlete candidates. Furthermore, an explainable AI (XAI) technique, Shapley Additive Explanations (SHAP), is applied to interpret model decisions and provide meaningful insights into key predictive factors. Experimental results demonstrate that Gradient Boosting achieves superior predictive performance (F1) across the 10-fold sets, with a mean value of 0.46. SHAP analysis reveals the critical importance of anthropometric measurements and social group features in influencing prediction outcomes. These findings collectively underscore the substantial potential of ML to revolutionize talent identification in sports while emphasizing the importance of model interpretability in fostering trust and acceptance of AIdriven decision-making processes.
Updating road network maps is essential for transportation services, as incomplete or inaccurate maps can lead to inefficiencies and diminish service quality. The online transportation industry generates vast amounts of GPS data as drivers navigate, which is valuable for mapping road networks and improving traffic management. However, since drivers do not cover all roads, satellite imagery plays a crucial role in identifying areas that are not mapped. By combining GPS data as labels with satellite imagery, the extraction of new road networks becomes more accurate. This research employs a deep learning Convolutional Neural Network (CNN) with the U-Net architecture for road segmentation, allowing for the identification of new paths. Two different encoders are tested in this research: Inception-ResNet-V2 and a pure U-Net encoder. The Inception-ResNet-V2 encoder achieves an accuracy of 91.3%, while the pure U-Net encoder achieves 90.7%. In terms of Dice Loss, the models record values of 0.051 and 0.08, respectively. The research highlights the effectiveness of different U-Net encoders in road network segmentation. With high accuracy and low Dice Loss, this approach provides a reliable method for automatically updating road maps. It has potential applications in navigation systems, urban planning, and AI-driven intelligent transportation systems.
This research evaluates the performance of Artificial Neural Network (ANN) models in forecasting temperature at Djuanda Airport, comparing them with the traditional Autoregressive Integrated Moving Average (ARIMA) model and a hybrid ARIMA–ANN approach. Although statistical models such as ARIMA are widely applied, their capacity to capture nonlinear dynamics in tropical climate conditions is limited, particularly when the data exhibit irregular fluctuations that linear models cannot adequately represent. Forecasting temperatures in tropical airport settings, which is crucial for flight planning, operational safety, and the reliability of aviation operations, remains relatively underexplored. This gap underscores the importance of alternative modeling techniques that can effectively address nonlinear relationships. Using one year of observed data, the models are evaluated with three accuracy metrics: Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Squared Error (RMSE). The ANN model achieves the lowest error values (MAE 0.7630, MAPE 2.7067%, RMSE 1.0074) compared to both ARIMA and hybrid approaches. The metrics and the testing graph collectively indicate that ANN has a stronger ability to capture nonlinear temperature dynamics in tropical contexts. Nonetheless, the findings must be interpreted with caution due to the limited dataset and single case study. These limitations highlight the need for extended data and alternative architectures to improve forecasting accuracy and strengthen support for safer aviation operations.
Drowsiness is a problem that needs to be addressed to improve road safety. To minimize this safety issue, driving-monitoring systems have been implemented in current car models, and electrocardiography (ECG) is one of the most commonly used driving monitoring techniques. ECG data are modeled using a deep neural network, including a Bidirectional Gated Recurrent Unit (Bi-GRU). However, the accuracy for classifying Wake-Sleep is under 80% and Wake-NREM-REM reaches less than 68%. To address this issue, ECG data from the MESA and SHHS datasets are modeled using a combination of a Convolutional Neural Network (CNN) and a Bi-GRU, referred to as CNN-GRU. This model incorporated Batch Normalization and RMSProp to achieve improved accuracy in classifying drivers' conditions. It operates in two computing sectors: cloud computing (Google Colaboratory, also known as Colab) and edge computing (utilizing an AMD Ryzen 5 4600H processor laptop). Those computing sectors focused on a case where no internet connectivity occurred to process the classification. Those classifications achieved accuracy rates of 82.88% and 81.78% for Wake-Sleep classification in cloud- and edge-computing, respectively. Additionally, it achieved 71.01% (Colab) and 68.85% (edge-computing) accuracy in Wake-NREM-REM classification. This result indicates that CNN-GRU achieved better performance, surpassing the previous Bi-GRU model, which only achieved 80.42% (Colab) and 76.2% (edge-computing) for Wake-Sleep, and 68.85% (Colab) and 66.43% for Wake-NREM-REM.
The behavior of petroleum reservoirs is inherently complex, making it challenging to determine their performance for both single-fluid and multiphase production systems. To accurately estimate the recovery reserves of a reservoir, a comprehensive understanding of its geometry and internal flow characteristics is essential. Numerical simulation serves as a fundamental tool for reservoir engineers, offering an efficient and reliable method to predict reservoir mechanisms, evaluate pressure variations, and estimate in-place hydrocarbon yield. This study employs mathematical modeling concepts and numerical techniques to analyze the dynamic behavior of petroleum reservoir systems. A flow model based on Partial Differential Equations (PDEs), specifically the diffusivity equation for unsteady-state fluid flow in porous media, is developed and applied. The diffusivity equation is discretized and solved mathematically using the explicit finite difference method to approximate pressure distribution over time and space. The primary objective of this research is to investigate and analyze the pressure distribution that governs reservoir performance under varying conditions. Sensitivity analyses are conducted to evaluate the influence of grid spacing, time step, hydraulic diffusivity, and boundary conditions on pressure reservoir behavior within a Cartesian grid for a one-dimensional, single-phase reservoir. The findings are expected to provide insight into the relationship between reservoir properties and fluid dynamics, supporting improved prediction of reservoir behavior. Ultimately, this research contributes to the optimization of petroleum production strategies and enhances the understanding of reservoir engineering processes through quantitative simulation.
As K-pop continues to dominate global music charts, understanding the factors behind the success of songs has become increasingly essential. This study explores how musical elements and popularity indicators reveal patterns among topperforming songs. A total of 57 songs nominated for the 2024 Song of the Year category were grouped using hierarchical cluster analysis. The genre variable was consolidated into six broader categories and converted into numerical labels. All variables are normalized using the Min-Max normalization method before clustering. The data includes musical elements such as genre, tempo, danceability, energy, and happiness, as well as popularity indicators like YouTube views and Spotify streams. The analysis employs single, complete, and average linkage methods. Among these, the average linkage method yields the best results, with an agglomerative coefficient value of 0.8167. Seven distinct clusters are identified: Cluster 1 features R&B and hip-hop styles with varied energy and rhythms; Cluster 2, the largest group, includes high-energy pop, hip-hop, and dance-pop tracks that are popular on streaming platforms; Cluster 3 contains indie and experimental tracks; Cluster 4 emphasizes high-energy stage performances; Cluster 5 is an outlier with experimental traits; Cluster 6 highlights R&B and funk with global appeal; and Cluster 7 includes emotional OSTs and ballads with slower tempos. By combining musical elements and popularity indicators, this research uncovers patterns of success in K-pop songs. These findings offer actionable insights for artists, producers, and marketers, providing a datadriven reference for creating music that resonates with modern audience preferences.
The in-situ soil infiltration test using A Double Ring Infiltrometer (DRI) apparatus can be conducted in the field according to DIN 19682-7 standards and procedures. As required by these standards, the traditional paper-based measurement form can be replaced with a new application developed to meet standard requirements. The DRI apparatus consists of two concentric rings placed in the soil, filled with water, while the outer ring maintains a constant water level. The water level drop in the inner ring is observed and recorded at regular intervals. The infiltration rate can be calculated for each interval by measuring the change in water height over time. This new application facilitates the automatic calculation of both the actual soil infiltration rate and the Horton soil infiltration model. Comparison tests between the application results and Excel calculations have yielded similar outcomes. The goal of this research is to develop a mobile web-based application for recording data and calculating soil infiltration measurements using the DRI method. The research methodology involves transforming the measurement procedure into a concept, designing the application, and then implementing that design. By replacing the paper-based process, this application will enhance the efficiency, accuracy, and flexibility of soil infiltration measurement projects in various locations. Furthermore, the data will be stored in the cloud, allowing for crowdsourced infiltration data collection and monitoring from any location, including the office.
The research aimed to optimize the quality attributes of Piper retrofractum Vahl.—piperine content, color brightness, and water content—using Partial Least Squares Regression (PLSR) to evaluate the pretreatment effects with fruit peel infusions and drying conditions. The research urgency lied in addressing the challenges of achieving consistent product quality while promoting sustainable food processing practices. Around 30 samples of Piper retrofractum Vahl. were subjected to varying pretreatment concentrations, soaking durations, drying durations, and peel types (orange and pineapple). The PLSR model was employed to identify key factors influencing the quality attributes and assess predictive performance based on Root Mean Squared Error (RMSE) and Coefficient of Determination (R²) values. As a result, the PLSR model explains 43.22% of the variance in piperine content, highlighting the importance of shorter soaking durations and higher pretreatment concentrations in preserving piperine levels. For water content, the model captures 75.08% of the variance, emphasizing the critical role of drying duration in reducing moisture. However, the color brightness model explains only 18.5% of the variance, indicating the need to explore contributing factors further. The research introduces the innovative use of fruit peel-infused water as a sustainable pretreatment method, contributing to eco-friendly food processing practices and offering practical insights into optimizing production for improved product quality. The findings underscore the importance of balancing pretreatment and drying parameters to address inconsistencies in quality while promoting sustainability. Future research should expand experimental conditions, integrate additional variables, and explore advanced modeling techniques to enhance predictive accuracy and product quality.
The research sought to data mine the financial literacy of tertiary students to evaluate and pinpoint deficiencies in their financial knowledge, measure the degree of financial goal-setting and budgeting practices, and determine their primary sources of financial advice. The data were gathered through validated questionnaires and disseminated through surveys. The research focused on tertiary students with 316 valid responses for analysis. The research used data mining techniques under the Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology to identify students’ financial literacy. Furthermore, a comprehensive analysis using Ms Excel, Statistical Package for Social Sciences (SPSS), and Waikato Environment for Knowledge Analysis (WEKA) reveals significant disparities in financial literacy among various fields of study and course levels. The investigation highlights essential financial behaviors, such as credit card utilization, saving patterns, and budgeting strategies, while revealing deficiencies in formal financial education. The analysis highlights the necessity for specialized financial literacy initiatives in educational programs to bridge knowledge deficiencies and encourage proficient budgeting and goal-setting techniques. The results offer practical guidance for educators, policymakers, and higher education institutions to improve students’ financial well-being, in line with Sustainable Development Goals (SDGs) focused on poverty alleviation and economic development. The research advocates for financial literacy programs in the school curriculum and emphasizes enhancing student participation in workshops. Higher education institutions must provide well-structured financial advice and support services. Lastly, Future studies should delve deeper into socioeconomic factors to improve predictive models and intervention strategies.
Small and Medium Enterprises (SMEs) have experienced rapid growth, contributing approximately 95% to the global economy, 60% to global employment, and 50% to global GDP. This growth is accompanied by significant challenges, with approximately 70% of SMEs failing within the first three years, primarily due to poor inventory management. It emphasizes the crucial role of accurate demand forecasting for SMEs, particularly in the retail sector, where time series at various levels of hierarchical structure exhibit different scales and display diverse patterns. However, most existing research on demand forecasting for SMEs focuses on a single hierarchical level—either bottom, middle, or top—without addressing the entire hierarchy. The research sought to address this gap by forecasting across all hierarchical levels and evaluating different reconciliation techniques to generate coherent and accurate forecasts for multiple products in retail SMEs. The ETS state space model was used as the base forecasting model. This model was widely recognized as a benchmark in forecasting competitions. The reconciliation methods assessed were Bottom-Up, Top-Down based on historical proportions (average proportions), Top-Down based on forecast proportions, and Minimum Trace (MinT) (Ordinary Least Squares (OLS), OLS Non-Negative (OLS Non-Neg), Weighted Least Squares (WLS), and WLS Non-Negative (WLS Non-Neg)). The evaluation results show that the OLS Non-Negative method, with an average SMAPE value of 35.335%, produces more accurate reconciliation than other methods. In addition, this method also outperforms the base model with an increase in accuracy of 13%.
The research was conducted to determine the descriptive statistics of suicide cases and classify suicide cases based on the attributes of victims who made suicide attempts. The research design used was a quantitative method in the form of exploratory research using the Decision Tree method. The research novelty was applying the Decision Tree method with the Best Subset approach. The research data sources were obtained from online mass media news such as DetikJatim and DetikJateng for suicide attempt cases from January 2022 to July 2024. The research finds significant differences in the number of suicide attempts in East Java and Central Java, with Surabaya, Malang, Blitar, Semarang, and Klaten recording higher numbers. The findings show that males more often attempt suicide, while females more often experience failed attempts. Young adults (20−39 years) record the highest rate, and hanging is the most common method. Unknown mental disorders and depression are the main risk factors, with many attempts occurring without rescue. The implication is that improving emergency response systems and mental health services is essential. The research recommends strengthening mental health and social support for older adults and those under stress. Then, enhancing rapid rescue efforts with comprehensive psychological interventions is essential for suicide prevention. The originality of the research lies in the use of a Decision Tree with the Best Subset approach to identify suicide patterns based on risk factors and methods used.
Clustering groups aims to ensure similarity within clusters and disparity between them. The research evaluated the Fuzzy C-Means method’s effectiveness in clustering large datasets containing outliers, focusing on the 2021 Village Potential data from Bengkulu Province. The dataset, comprising 1,514 observations from villages and urban villages, provided a comprehensive resource for understanding regional development. Outliers, a common challenge in cluster analysis, were detected using univariate and multivariate methods, revealing substantial variability. PCA was applied, improving clustering quality to address multicollinearity among variables. In the results, the fuzzifier (w) parameter in the FCM method plays a crucial role in controlling the degree of membership for data points in clusters, which can potentially reduce the impact of outliers, enhancing clustering robustness and accuracy. The FCM method effectively produces clusters with high intra-cluster homogeneity and inter-cluster heterogeneity. Using the Elbow method, three optimal clusters are identified. Cluster 1, dominated by villages in Bengkulu City, is the most advanced, with superior infrastructure and services, but the fewest villages business units, necessitating economic empowerment. Cluster 2, comprising villages in North Bengkulu Regency, demonstrates moderate development but suffers from poor transportation access, requiring improvements to support socio-economic activities. Cluster 3, dominated by villages in Kaur Regency, is the least developed, with limited basic services and infrastructure, highlighting the need for substantial investments in governance and essential services. These findings provide actionable insights for village development in Bengkulu Province, supporting targeted policies tailored to each cluster’s unique characteristics.
Prioritizing school building maintenance solely based on structural damage often leads to inefficient budget allocation and fewer beneficiaries. The research introduced an integrated Cost-User Effectiveness Ratio (CUER) to establish maintenance priorities by combining three critical factors: damage severity, maintenance costs, and the number of affected students. The CUER formulation employed the Geometric Mean or the root mean multiplication of the cost effectiveness and user effectiveness ratio to balance these factors systematically. The methodology encompassed several steps, including damage assessment and calculation of component importance weights using the Analytical Hierarchy Process (AHP), to determine integrated damage levels, costs, and student weights. These inputs were subsequently used to generate priority rankings of schools requiring maintenance. As a result, the case study in Wonogiri Regency illustrates the superiority of the proposed method over the conventional method. While the conventional approach prioritizes 27 schools benefiting 2,442 students, the CUER approach prioritizes 33 schools benefiting 2,957 students, demonstrating increased efficiency and broader impact. The CUER-based model presents a systematic and equitable solution to prioritize school building maintenance, ensuring the optimal allocation of resources and maximizing benefits within existing budgetary constraints. This innovative approach addresses current challenges in maintenance planning and offers significant implications for improving the management of educational infrastructure.
