Publication Search

Background: The development of modern manufacturing systems requires production scheduling strategies that not only improve productivity but also optimize energy utilization. Multi-machine production systems with job-shop configurations exhibit high complexity due to dynamic interactions between machines, job queues, and varying processing times, making conventional scheduling methods less effective in handling changing operational conditions. Objective: This study aims to develop and evaluate a reinforcement learning based production scheduling approach to improve production efficiency while reducing energy consumption in multi-machine manufacturing systems. Methods: This research employs a job-shop based multi-machine production simulation model as the experimental environment. The scheduling problem is formulated as a Markov Decision Process, enabling the implementation of reinforcement learning algorithms, namely Q-learning and Deep Q-Network, to learn optimal scheduling policies through interaction with the simulation environment. Energy consumption parameters are incorporated into the reward function so that the learning agent can consider energy efficiency in the scheduling decision-making process. System performance is evaluated using three main metrics, namely energy consumption, throughput, and makespan. Results: The experimental results show that the reinforcement learning based scheduling approach achieves better performance compared to conventional scheduling methods, resulting in lower energy consumption, higher job completion rates, and shorter production completion times within the multi-machine manufacturing system.

Jambi Province is one of the largest natural rubber producing regions in Indonesia; however, rubber factories under GAPKINDO Jambi still face productivity issues, particularly the gap between production capacity and actual output, and productivity assessment that is still conducted manually by GAPKINDO Jambi. This study employs Decision Tree, Random Forest, KNN, and SVM algorithms within a structured pipeline involving preprocessing, feature selection, standardization, data balancing using SMOTE, and hyperparameter tuning. The proposed solution applies productivity level classification both individually and through paired combinations (ensemble voting). The results show that the Decision Tree + Random Forest model achieves the best performance with an accuracy of 0.84 and an F1-score of 0.83, confirming the effectiveness of ensemble methods in supporting productivity improvement decisions.

AImbalanced data remains a significant issue in heart disease classification using machine learning, as it tends to cause models to overestimate the majority class while ignoring minority classes with high clinical value. This can lead to a decrease in accuracy and the model's ability to accurately detect disease cases. Therefore, this study aims to assess the effectiveness of oversampling techniques, namely Random Oversampling and Synthetic Minority Oversampling Technique (SMOTE), in improving the performance of the K-Nearest Neighbors (KNN), Naive Bayes (NB), and Random Forest (RF) algorithms. The dataset used comes from Kaggle and consists of 918 data sets with 12 attributes representing patient information related to heart disease prediction. The research stages include data preprocessing, baseline model testing, and re-evaluation using the two oversampling methods. Experimental results show that oversampling can improve the performance of all algorithms. KNN achieved the best results with SMOTE, with an accuracy of 72.98% and an F1-score of 75.39%. In the Naive Bayes algorithm, both oversampling techniques produced relatively stable performance, with the highest F1-score of 73.56% using SMOTE. Meanwhile, Random Forest showed the most optimal performance when combined with Random Oversampling, with an accuracy of 79.19% and an F1-score of 81.51%. These findings confirm that the success of data balancing techniques is strongly influenced by the characteristics of the classification algorithm used, and provide a practical contribution in determining strategies for handling imbalanced data in health research.

This study aims to classify customer payment methods at 17 Coffee & Eatery using machine learning algorithms, namely Naïve Bayes and Support Vector Machine (SVM). The increasing use of digital and non-cash payments has generated large volumes of transaction data that are rarely analyzed optimally, even though such data contain valuable information for business decision making. This research used secondary transaction data collected from January to March 2025, consisting of 10,147 transaction records. The dataset included several attributes such as order time, payment time, transaction type, total sales, number of items, and payment method. Data preprocessing was performed through data cleaning, feature engineering, normalization, and label encoding before being divided into training and testing sets with an 80:20 ratio. The Naïve Bayes and SVM models were then trained and evaluated using accuracy, precision, recall, F1-score, and ROC–AUC metrics. The results show that both algorithms were able to classify payment methods effectively, but SVM achieved higher accuracy and more stable performance than Naïve Bayes. These findings indicate that SVM is more suitable for handling complex and heterogeneous transaction patterns. The implementation of machine learning for transaction classification can support more efficient financial management and data-driven decision making for small and medium enterprises in the culinary sector.

In recent years, with the rapid development of artificial intelligence, many innovative changes have been made in the field of intelligent mobile robot development. In the field of control and navigation of mobile robots, learning-based methods have many advantages over traditional ones. The study of mobile robot control methods using deep reinforcement learning is a remarkable area in the development of mobile robots that must operate in dynamic environments. In the previous studies, the proposed robot control algorithms using deep reinforcement learning are mostly based on the given target point and obstacle information, the robot path planning is performed, and the corresponding control is based on the obtained path. The DDPG-based method is a typical example. However, in dynamic environments, DRL based robot path planning requires a state of target point and obstacles information, which leads to a large amount of computation, resulting in extremely long convergence time and even non-convergent cases. In this paper, we propose a new method for mobile robot control in dynamic environment that solves the dimensional problem by extracting the features of the configuration of obstacles using autoencoder and learning the DDPG algorithm based on the obtained features. Simulation results show that the proposed algorithm can effectively solve the mobile robot control problem in dynamic environment.

The growing demand for energy-efficient and intelligent thermal systems has driven significant advancements in adaptive compressor design. This paper presents a comprehensive literature review on the development of AI-based compressor systems, with a specific focus on enhancing efficiency under partial-load conditions and optimizing the utilization of residual energy. Through the synthesis of five recent high-impact studies (2020–2025), we examine the application of deep reinforcement learning (DRL), hybrid evolutionary algorithms, and neural network surrogate modeling in compressor optimization. Key findings indicate that model-based DRL combined with surrogate CFD can achieve up to 8% efficiency gains at off-design conditions. Hybrid approaches integrating Genetic Algorithms (GA) with DRL reduce optimization time by 30% while improving pressure ratios. Neural network surrogates provide high-speed, real-time performance predictions with less than 1% error, enabling mass iterative design. Furthermore, intelligent load classification using radial basis function networks (RBFN) allows adaptive response to varying operating conditions with over 95% accuracy. Collectively, these methods form a framework for intelligent, self-optimizing compressor systems capable of real-time adaptation and energy recovery. The results suggest that AI-enhanced adaptive compressors represent a transformative direction for energy-sensitive sectors, including HVAC, power generation, and sustainable industry.

The rapid growth of open-source software (OSS) in machine learning (ML) has intensified the need for reliable, automated methods to assess project quality, particularly as OSS increasingly underpins critical applications in science, industry, and public infrastructure. This study evaluates the effectiveness of a diverse set of machine learning and deep learning (ML/DL) algorithms for classifying GitHub OSS ML projects as engineered or non-engineered using a SMOTE-enhanced and explainable modeling pipeline. The dataset used in this research includes both numerical and categorical attributes representing documentation, testing, architecture, community engagement, popularity, and repository activity. After handling missing values, standardizing numerical features, encoding categorical variables, and addressing the inherent class imbalance using the Synthetic Minority Oversampling Technique (SMOTE), seven different classifiers—K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), XGBoost (XGB), Logistic Regression (LR), Support Vector Machine (SVM), and a Deep Neural Network (DNN)—were trained and evaluated. Results show that LR (84%) and DNN (85%) outperform all other models, indicating that both linear and moderately deep non-linear architectures can effectively capture key quality indicators in OSS ML projects. Additional explainability analysis using SHAP reveals consistent feature importance across models, with documentation quality, unit testing practices, architectural clarity, and repository dynamics emerging as the strongest predictors. These findings demonstrate that automated, explainable ML/DL-based quality assessment is both feasible and effective, offering a practical pathway for improving OSS sustainability, guiding contributor decisions, and enhancing trust in ML-based systems that depend on open-source components.

The rapid growth of web technologies and online services has increased the exposure of web applications to cyber threats such as Cross-Site Scripting (XSS) and SQL Injection (SQLi). Conventional rule-based mechanisms, such as Web Application Firewalls (WAFs), often fail to detect emerging attack patterns. To address this, Machine Learning (ML) and Deep Learning (DL) have emerged as adaptive approaches for enhancing web attack detection. This study performs a Systematic Literature Review (SLR) following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines to analyze recent ML/DL-based detection methods. Of the 263 retrieved studies, 15 met the inclusion criteria for detailed review. The findings reveal that Random Forest (RF), Support Vector Machine (SVM), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) are the most applied algorithms. At the same time, recent works emphasize Transformer-based and hybrid ML–DL models. These approaches achieved robust performance (accuracy 85–97%, F1-score >90%) but still face challenges in dataset representativeness, class imbalance, and computational cost. This review highlights future research directions in Explainable Artificial Intelligence (XAI), Federated Learning (FL), and adversarial robustness to develop more efficient and trustworthy web attack detection systems.

Millions of people suffer from malaria, one of the most serious parasitic diseases that threatens human life and causes high rates of morbidity and mortality, particularly in tropical and subtropical regions. Traditional diagnostic methods, such as blood smear examination, which can be performed using a microscope, face many challenges due to the inaccuracy of manual analysis and the reliance on individual skills. Therefore, the use of machine learning or deep learning algorithms to automate malaria detection offers promising solutions to improve accuracy, reduce diagnosis time, and enhance scalability. In this paper, a multi-class convolutional neural network (CNN)-based model is designed to classify cells infected with Plasmodium falciparum (P. falciparum) and Plasmodium vivax (P. vivax) and uninfected cells from blood smears, as most severe cases and deaths are caused by P. falciparum and P. vivax. This is achieved by building and training a CNN from scratch, rather than using transfer learning from pre-trained models. The proposed network was trained and tested on the Kaggle dataset, which consists of 27,558 images of infected and uninfected individuals. These images were divided into 13,779 images of uninfected individuals, 6,890 images of individuals with P. falciparum malaria, and 6,889 images of individuals with P. vivax malaria. The images were preprocessed using several operations, including blurring, denoising, and morphological processing. The proposed model achieved the best evaluation accuracy when compared with other deep learning algorithms, with an accuracy rate of 96.5%, a sensitivity rate of 95%, a specificity rate of 97.6%, and an F1-score rate of 96.5%. These results demonstrate the effectiveness of the proposed model as a tool to assist clinicians in malaria diagnosis, reducing reliance on manual analysis.

Typhoid fever is an infectious disease caused by the bacterium Salmonella typhi, commonly found in developing countries, including Indonesia. Prompt and accurate treatment is crucial to prevent serious complications in patients. One way to assist in diagnosing typhoid fever is by applying machine learning methods to classify patient data. The Naive Bayes method is one of the machine learning algorithms frequently used in medical data classification due to its strong ability to handle large and complex datasets. This article discusses the application of the Naive Bayes method for classifying typhoid patient data at Rantauprapat General Hospital (RSUD Rantauprapat). By utilizing medical data that includes clinical symptoms, laboratory test results, and patients’ medical histories, the Naive Bayes model can provide fairly accurate predictions regarding the likelihood of a person having typhoid fever. The research findings indicate that Naive Bayes is reliable in predicting typhoid diagnoses with adequate accuracy, thereby supporting healthcare professionals in making faster and more precise decisions. It is expected that the implementation of this method can accelerate the diagnostic process and improve the quality of healthcare services at RSUD Rantauprapat, as well as in other regions.

Dengue Fever (DF) continues to be a major public health threat in Indonesia, especially in urban areas with high population density, such as Purwokerto City. This study aims to develop a predictive model to identify high-risk areas for DF outbreaks by integrating Machine Learning (ML) algorithms and Geographic Information Systems (GIS). The research utilizes historical dengue case data, meteorological parameters (rainfall, temperature, humidity), and population density as predictive variables. Three ML classification algorithms—Naïve Bayes, Logistic Regression, and Support Vector Machine (SVM)—were implemented to develop risk prediction models. Extensive data preprocessing, feature selection, and spatial integration were applied to ensure model robustness. The results show that the SVM model outperformed other methods, achieving the highest accuracy, precision, recall, and F1-score in classifying dengue risk zones. Risk maps generated through GIS visualization successfully identify priority areas for targeted interventions. The novelty of this research lies in the combination of local epidemiological data, multi-algorithm comparison, and geospatial mapping to improve early warning systems for DF in Purwokerto. This integrated approach is expected to support more effective prevention strategies and enhance public health preparedness.

Learners are continually faced with choosing appropriate courses or making career choices due to increased educational opportunities. The emergence of machine learning-based course and career recommender systems has the potential to address this issue, offering personalized course recommendations tailored to individual learning pathways, preferences, and learning history. The optimization and feature engineering techniques and practical deployment environments have not been collectively examined in the previous research, despite the significant advancements in this area of research. Furthermore, previous research has rarely synthesized how these technical components help students choose appropriate courses and careers. This systematic review was carried out to investigate the current state of machine learning-based course and career recommender systems, focusing on key elements, such as primary data sources, feature engineering methods, algorithms, optimization techniques, evaluation metrics, and the environments where the existing course recommendation models are deployed. The PRISMA method for conducting a systematic review was used to choose studies that met the requirements for inclusion and exclusion. The study findings show significant reliance on interpretable and traditional machine learning algorithms, such as K-Nearest Neighbor and Random Forest, to develop recommender models. Feature engineering remains basic, as most studies rely on normalization, while optimization processes are often underreported. Also, evaluation metrics varied widely, impeding comparability, while most of the recommender models are deployed in an e-learning environment, leaving the traditional learning environment underrepresented. Furthermore, the study findings identified issues including data sparsity and diversity, data security and privacy, and changes in learner preferences that may have an impact on the performance of recommender systems while recommending further studies to make use of standardized optimization methods, and automated domain-informed feature engineering frameworks, benchmark and annotated datasets in developing models the gives priority to learners’ success and educational relevance.

The COVID-19 pandemic has exposed fundamental vulnerabilities in global supply chain systems, such as over-reliance on single suppliers and a lack of operational visibility. This has highlighted the urgent need for a new approach to risk management—one that leverages smart technologies. Artificial Intelligence (AI) has emerged as a promising solution, thanks to its capabilities in predictive analytics and adaptive, data-driven decision-making in real time. This study aims to develop an AI-based predictive system framework to enhance the resilience of global supply chains in the face of post-pandemic disruptions. Using the Design Science Research (DSR) methodology, the research designs and evaluates a system that integrates algorithms such as LSTM, Random Forest, Natural Language Processing (NLP), and Reinforcement Learning. It also applies a federated learning approach to ensure data privacy among supply chain partners. The study analyzes over 12,000 data entries from diverse sources, including IoT devices, weather data, demand trends, and social media. The system's effectiveness is evaluated through a combination of quantitative methods (PLS-SEM analysis on 103 respondents) and qualitative methods (interviews with 12 industry executives). The findings show that AI-driven predictive analytics significantly improve supply chain resilience (β = 0.67; p < 0.001), with demand forecasting accuracy increasing by up to 40% and delivery times reduced by 30%. Conceptually, the study contributes by designing a resilient model that integrates real-time visibility, adaptability, and cross-organizational collaborative learning. Unlike traditional approaches focused solely on automation, this framework offers a more holistic solution, addressing key gaps in the literature. The implication is clear: AI is becoming a strategic asset in building sustainable, resilient supply chains amid ongoing global uncertainty.

The increasing use of mobile devices has increased the risk of data theft, posing significant security challenges for individuals and organizations. This study proposes an early detection system for data theft on mobile devices using machine learning algorithms. The system is designed to identify suspicious patterns in application usage, network access, and CPU/memory activity, providing early warnings to prevent potential data loss. By employing algorithms such as Random Forest, Support Vector Machine (SVM), and Convolutional Neural Network (CNN), the developed models demonstrated significant performance: CNN achieved the highest accuracy of 95.1%, with a precision of 94.2%, recall of 93.5% , F1-score of 93.8%, and AUC-ROC of 0.96. Random Forest and SVM also showed competitive performance with accuracy rates of 94.7% and 92.5%, respectively. These findings highlight the high potential of machine learning algorithms for real-time detection of data theft threats, providing adaptive protection against evolving cyberattack methods. This approach offers a promising solution to strengthen mobile device security frameworks and safeguard user data against increasingly sophisticated cyber threats.

Image processing is a branch of computer science that is growing rapidly and is widely used in various fields, including in security systems. Face identification is one of the main applications of image processing that aims to recognize and distinguish individual faces in a system. The methods used in face identification involve various techniques, such as facial feature detection, characteristic extraction, and classification using machine learning algorithms. This article discusses the application of image processing in a security system based on face identification and the technology used to improve the accuracy and reliability of the system. The results of the study show that the combination of deep learning algorithms with image pre-processing techniques can increase the success rate of face identification in security systems.

The interaction between genetic and environmental factors plays a crucial role in determining phenotypic traits in organisms. This study aims to analyze these interactions using computational approaches, including statistical models and machine learning algorithms. The data used include genetic factors (genotypes) and simulated environmental factors. Results indicate that machine learning models such as Random Forest can detect interaction patterns with high accuracy, as demonstrated by significant R² values. Additionally, heatmap visualizations provide deeper insights into the non-linear effects of genetic-environment interactions. This study highlights the potential of computational methods in exploring complex interactions, with broad applications in health, agriculture, and biotechnology.

Introduction: Achieving carbon neutrality in industrial systems is essential for mitigating climate change and promoting sustainability. The increasing demand for energy optimization and carbon emission reduction has driven the development of advanced technologies, particularly hybrid machine learning (ML) models. These models, combining ensemble learning and reinforcement learning (RL), offer significant promise in optimizing industrial processes, reducing energy consumption, and improving environmental performance. This study explores the application of hybrid ML models in achieving carbon neutral goals through dynamic process optimization and energy control in industrial settings. Literature Review: Hybrid ML models integrate different machine learning techniques to handle complex and dynamic environments effectively. Ensemble learning methods, such as boosting, bagging, and stacking, combine multiple algorithms to improve predictive performance and robustness. Reinforcement learning (RL), on the other hand, enables real time decision making and adaptation based on trial and error interactions with the environment. In energy optimization, these models are used to reduce energy intensity and carbon emissions, enhancing overall operational efficiency. Previous studies have demonstrated the effectiveness of ML models in energy management, but challenges such as data quality, model integration, and computational complexity remain. Materials and Method: The study applies hybrid ML models combining ensemble learning and RL to optimize energy consumption and minimize carbon emissions in industrial processes. Data from real time sensors and operational parameters are used to train the models. The ensemble learning component improves the accuracy of energy predictions, while RL ensures dynamic process adjustments in response to fluctuating energy demand. The models were tested in various industrial settings, including manufacturing processes, smart grids, and microgrid systems. Performance metrics such as energy efficiency, carbon emissions reduction, and operational costs were evaluated to assess the effectiveness of the models.  Results and Discussion: The hybrid ML models achieved significant reductions in energy intensity (15-20%) and carbon emissions (18-25%). The real time adaptability of the RL component allowed the models to adjust energy consumption patterns dynamically, improving energy efficiency and reducing waste. The models demonstrated their ability to adapt to varying operational conditions, ensuring optimal energy use. A cost-benefit analysis showed that the hybrid models provided substantial energy savings and reduced operational costs, with a return on investment (ROI) of 30-35% within the first year of deployment. However, challenges such as computational complexity and data quality issues were identified, highlighting the need for further refinement in model development.

Modernization of education in Islamic boarding schools is an important need to maintain the relevance of learning in the digital era. Salafiyah Islamic boarding schools, which are synonymous with the traditional approach through the study of the Yellow Book, face the challenge of adapting technological developments without losing their scientific character. This research aims to develop a model of curriculum modernization through the integration of deep learning-based digital technology  to improve the effectiveness of learning and the quality of academic evaluation. The methods used include literature studies, observation of the learning process, and trial implementation of digital learning tools equipped with deep learning algorithms  for student ability analysis. The results of the study show that the application of deep learning technology  is able to speed up the competency mapping process, provide automatic feedback on reading and comprehension of classical texts, and assist teachers in monitoring the development of students more accurately. This integration also increases interest in learning and strengthens the mula curriculum system without shifting the traditional values of the Islamic boarding school. Thus, technology-based modernization can be a strategic solution to bridge the classical approach and the needs of 21st century learning in Salafi Islamic boarding schools.

The integration of solar powered IoT irrigation systems in precision agriculture offers a sustainable solution to address water scarcity and enhance crop productivity. By leveraging real time data from soil sensors, weather APIs, and machine learning algorithms, these systems optimize irrigation schedules and improve water use efficiency. This research explores the potential of integrating renewable energy sources, such as solar power, with edge computing in smart irrigation systems to promote sustainable agricultural practices. The study aims to evaluate the performance of the proposed system in terms of water savings, crop yield, energy efficiency, and adaptability to varying climate conditions. Literature Review: Previous studies highlight the importance of smart irrigation systems in reducing water waste and improving crop yield through real time monitoring and automated decision making. However, existing systems often lack the integration of renewable energy and edge computing, which are critical for ensuring sustainability and operational efficiency in rural agricultural settings. The combination of renewable energy with IoT devices offers a promising solution to reduce energy costs and carbon emissions, while edge computing enhances real time data processing, ensuring prompt and accurate irrigation adjustments. Materials and Method: The proposed system integrates solar powered IoT devices, soil moisture sensors, weather data APIs, and edge computing devices to manage irrigation. Machine learning algorithms and evapotranspiration models are used to predict irrigation needs and optimize scheduling based on real time data. The system's performance is evaluated through metrics such as water savings percentage, crop yield improvements, and energy consumption, with a comparative analysis against traditional irrigation methods. Results and Discussion: The results indicate that the system successfully reduces water usage by 30% to 40%, increases crop yield by 25%, and operates with energy autonomy, powered entirely by solar energy. The system's adaptability to varying climate conditions ensures optimal crop growth, even under environmental stresses. The integration of renewable energy and edge computing significantly enhances the sustainability and efficiency of irrigation systems.

Cybersecurity is continuously challenged by increasingly sophisticated and dynamic cyber-attacks, necessitating advanced adaptive defense mechanisms. Deep Reinforcement Learning (DRL) has emerged as a promising approach, offering significant advantages over traditional intrusion detection methods through real-time adaptability and self-learning capabilities. This paper presents an advanced adaptive cybersecurity framework utilizing five prominent DRL algorithms: Deep Q-Network (DQN), Proximal Policy Optimization (PPO), Twin Delayed DDPG (TD3), Soft Actor-Critic (SAC), and Asynchronous Advantage Actor-Critic (A3C). The effectiveness of these algorithms is evaluated within complex, realistic simulation environments using live-streaming data, emphasizing key metrics such as accuracy (AUC-ROC), response latency, and network throughput. Experimental results demonstrate that the SAC algorithm consistently achieves superior detection accuracy (95% AUC-ROC) and minimal disruption to network performance compared to other approaches. Additionally, A3C provides the fastest response times suitable for real-time defense scenarios. This comprehensive comparative analysis addresses critical research gaps by integrating both traditional and novel DRL techniques and validates their potential to substantially improve cybersecurity defense strategies in realistic operational settings.