Publication Search

High-dimensional clinical data exhibit complex and non-linear relationships among patient attributes, where outcomes are often influenced by feature interactions rather than isolated variables. However, many existing machine learning models prioritize predictive performance while providing limited interpretability and insufficient insight into interaction structures. This study aims to address this limitation by developing an interpretable and robust framework for feature interaction mining in clinical data. We propose a hybrid tree–neural modeling framework that explicitly captures and ranks feature interactions while maintaining stable predictive performance. Tree-based ensemble models are employed to identify non-linear interaction patterns, while neural representations enhance learning flexibility and generalization. The framework integrates interaction importance analysis, cross-validation–based stability assessment, and evaluation across multiple data splits to ensure robustness and interpretability. Experiments conducted on a real-world high-dimensional clinical dataset demonstrate that the proposed approach achieves consistent predictive performance, with AUC values ranging from 0.628 to 0.641 across five cross-validation folds (mean AUC ≈ 0.633). Performance remains stable under varying train–test splits, indicating strong generalizability. Interaction analysis reveals that a small number of dominant feature interactions—such as age combined with length of hospital stay and medication count combined with diagnostic information—consistently contribute to model predictions, appearing in over 80% of validation folds. Ablation studies further confirm that removing interaction-aware components leads to noticeable performance degradation, highlighting their importance.  In conclusion, this study demonstrates that explicit feature interaction modeling enhances interpretability, stability, and generalization in clinical prediction tasks. The proposed hybrid framework provides a reliable foundation for developing trustworthy and transparent clinical decision-support systems

This study presents a comprehensive comparative analysis of four traditional machine learning algorithms Decision Tree, Random Forest, K-Nearest Neighbors, and Support Vector Machine for Android malware detection using the preprocessed TUANDROMD dataset comprising 4,465 instances and 241 features representing both static and dynamic application characteristics. Motivated by the limitations of conventional signature-based and hybrid detection methods, especially in managing imbalanced datasets and detecting emerging malware variants, the study employed SMOTE to ensure balanced training data and fair model evaluation. The dataset was divided into 80% training and 20% testing subsets, and models were assessed using key performance metrics including accuracy, precision, recall, F1-score, and ROC AUC. The findings revealed that the proposed Random Forest model outperformed the other classifiers, achieving an accuracy of 0.993, precision of 0.992, recall of 1.000, F1-score of 0.996, and a near-perfect ROC AUC of 0.9998 surpassing state-of-the-art approaches. These results affirm the superior predictive capability, consistency, and robustness of the Random Forest algorithm in Android malware detection. The study concludes that base models, when integrated with class-balancing techniques, provide reliable and efficient malware detection across imbalanced datasets. For future research, the study recommends exploring advanced hybrid or ensemble frameworks that integrate Random Forest with deep learning architectures or other meta-heuristic optimization techniques to further enhance detection accuracy, adaptability, and resilience against rapidly evolving Android malware threats.

The rapid development of digital technology has led to an increased adoption of digital payment methods in online transaction-based businesses. However, in practice, failures and limitations in the implementation of digital payment systems still occur, potentially disrupting transaction processes and reducing customer convenience. Payment related obstacles may result in transaction cancellations and increase the risk of customer churn. This study aims to analyze the impact of failures and limitations in digital payment methods on customer churn using a classification-based approach. The data used in this research are secondary e-commerce customer data obtained from the Kaggle platform, including transaction information, payment methods, customer behavior, and historical transaction records. The research methodology consists of data preprocessing, time-based feature engineering, and classification modeling using logistic regression, decision tree, and random forest algorithms. Model performance is evaluated using accuracy, precision, recall, F1-score, and confusion matrix metrics. The results indicate that the decision tree model demonstrates superior capability in identifying churn customers compared to the other models, although it does not always achieve the highest accuracy. In addition to digital payment methods, other factors such as purchase value, transaction frequency, purchase timing patterns, and product return rates also influence customer churn. The findings highlight the importance of optimizing digital payment systems as part of customer experience enhancement strategies and customer retention efforts in online transaction–based businesses.

This study aims to analyze consumer purchasing behavior on e-commerce platforms using the Decision Tree algorithm as an easily interpretable classification method. The dataset used consists of 12,330 transaction records with 18 attributes representing visitor characteristics and user activities during interactions with the e-commerce platform. The research stages include data exploration to identify initial patterns, data preprocessing to handle missing values and class imbalance, splitting the data into training and testing sets, training the Decision Tree model, evaluating model performance, and visualizing the tree structure to analyze decision rules.The test results show that the Decision Tree model with a maximum depth of 3 achieves fairly good performance, with an average accuracy of 89.78%, precision of 69.82%, recall of 59.95%, and an F1-score of 64.51% for the buyer class. The visualization of the decision tree provides clear interpretation of the main attributes influencing purchasing decisions, thereby facilitating understanding for non-technical decision makers. Overall, this study demonstrates that the Decision Tree method is effective in modeling consumer purchasing behavior in e-commerce and can be utilized as a basis for data-driven business decision making, particularly in marketing strategies and improving sales conversion rates.

Decision-making under uncertainty is a major challenge in management, economics, and public policy, where outcomes cannot be accurately predicted due to limited information and environmental dynamics. This article conducts a systematic literature review of risk and probability approaches to decision-making under uncertainty, focusing on rational theory synthesis (such as expected utility theory, decision tree analysis, and Bayesian decision theory) and behavioral perspectives (prospect theory and heuristics). The review covers reputable literature from the last ten years to the present. The results show that the probabilistic approach provides a strong and adaptive rational framework, but has significant limitations due to cognitive biases such as loss aversion, overconfidence, and ambiguity aversion, which cause deviations from normative rationality. The integration of rational approaches with behavioral elements, through hybrid models, has proven to be more comprehensive and realistic for dealing with true uncertainty (Knightian uncertainty). These findings emphasize the need for a multidimensional decision-making paradigm in organizational and policy practices.

The retail industry generates an extremely large and continuously growing volume of transactional data along with the advancement of digital technology, thereby requiring sophisticated and systematic data analysis approaches to support effective and evidence-based business decision-making. This study aims to analyze retail sales data by utilizing the Retail Sales Dataset obtained from the Kaggle platform, which consists of 100,000 transaction records and broadly represents the characteristics of retail transactions. The main focus of this study is to classify product categories and predict customer segments, including the identification of high-spending customers (high spenders), based on demographic attributes such as age and gender, as well as various transaction-related features. The research methodology includes data preprocessing, label encoding, and feature engineering to generate additional variables, including Age_Group, Is_Holiday, and Spender_Group, which are expected to enhance the predictive capability of the models. Several machine learning algorithms, namely Decision Tree, Random Forest, and XGBoost, were implemented and evaluated to compare their respective performance. The experimental results indicate that multiclass product category classification achieves relatively low accuracy, ranging from 27% to 34%. These findings suggest the high complexity of retail data and highlight the need for further model optimization, class balancing techniques, and feature refinement to improve predictive performance in future studies.

 Autonomous robot navigation in dynamic and unstructured environments remains a critical challenge due to unpredictable obstacles, sensor uncertainty, and limited adaptability of traditional planning algorithms. Although conventional navigation methods such as graph-based, potential field–based, and sampling-based approaches have been widely adopted, their performance under real-time dynamic conditions is still constrained. This study aims to design and implement a comprehensive experimental framework to evaluate the effectiveness and limitations of conventional navigation algorithms for autonomous mobile robots operating in dynamic unstructured environments. The research adopts an experimental and comparative methodology by implementing A*, Dijkstra, Artificial Potential Field (APF), and Rapidly-Exploring Random Tree (RRT) algorithms in simulated static and dynamic scenarios. Performance is assessed using quantitative metrics including path length, computation time, success rate, collision rate, and path smoothness. The experimental results demonstrate that graph-based algorithms achieve high success rates and optimal path efficiency in static environments but exhibit limited adaptability to dynamic changes. APF offers fast computation but suffers from high collision rates due to local minima, while RRT shows better adaptability in dynamic environments at the cost of longer and less smooth paths. These findings confirm that conventional navigation methods are insufficient for robust autonomous navigation in highly dynamic and unstructured environments. The study highlights the necessity of adaptive and learning-based navigation frameworks, such as deep reinforcement learning, to enhance real-time decision-making, robustness, and autonomy in future robotic systems.

This study aims to develop an effective predictive model for identifying students at risk of academic dropout using the Decision Tree and Linear Regression algorithms. The data used are sourced from the public Kaggle dataset Students Dropout and Academic Success, which includes demographic, socioeconomic, and academic performance variables for each semester. The research method includes data preprocessing stages, such as data cleaning, label encoding for categorical variables, numeric feature normalization, and target class adjustment to focus on binary classification, namely Dropout and Graduate. The modeling process is carried out by comparing the performance of the two algorithms using evaluation metrics of accuracy, precision, and recall. The results show that the Decision Tree algorithm has superior performance compared to Linear Regression in mapping non-linear patterns in student data. Feature importance analysis revealed that the number of curricular units in the second semester and tuition payment status are the main predictors of dropout risk. These findings are expected to assist educational institutions in implementing early interventions to improve student academic success.  

Detection and identification of plant diseases is critical to the success and efficiency of agricultural production. Plant disease outbreaks are becoming more frequent throughout the world, and the presence of these diseases in cultivated plants has a significant impact on productivity. Therefore, researchers are focusing on developing effective and reliable plant disease detection methods. Thus, farmers can take advantage of early detection of this disease to minimize future losses. This article discusses machine learning approaches as well as decision trees, K-nearest neighbors, naive Bayes, support vector machines (SVM), and random forests for detecting coffee leaf diseases using leaf images. The above-mentioned classifications were researched and compared to determine the most suitable plant disease prediction model with the highest accuracy. Compared with other classification algorithms, the SVM algorithm achieves the highest accuracy of 99.75%. All the models trained above will be used by farmers to quickly identify and classify new diseases in images as a prevention strategy. As a preventive measure, farmers can detect and classify new diseases in images early.

Demam berdarah dengue merupakan tantangan kesehatan masyarakat yang terus berulang di wilayah tropis, termasuk Indonesia. Penelitian ini bertujuan untuk memprediksi jumlah kasus tahunan dengan memanfaatkan lima algoritma pembelajaran mesin, yaitu Regresi Linier, Decision Tree, Random Forest, Support Vector Machine (SVM), dan Neural Network. Data historis tahun 2017–2024 diolah menggunakan teknik windowing deret waktu untuk menghasilkan fitur lag yang sesuai bagi pembelajaran terawasi. Evaluasi kinerja dilakukan melalui metrik Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), serta koefisien determinasi (R²). Model Decision Tree menunjukkan performa paling unggul pada sebagian besar indikator. Prediksi untuk tahun 2025 mengindikasikan adanya peningkatan moderat jumlah kasus. Namun, rendahnya nilai R² pada seluruh model mengisyaratkan perlunya pendekatan multivariat yang lebih kompleks dengan mempertimbangkan faktor iklim, lingkungan, dan demografi. Hasil penelitian ini menegaskan pentingnya kualitas data dan pemilihan fitur yang tepat dalam peramalan epidemiologis guna mendukung perencanaan kesehatan yang lebih efektif.

Current advances in information technology have encouraged universities to utilize student academic data as a basis for decision-making, one of which is predicting academic achievement. This study aims to apply the C4.5 algorithm to develop a system for predicting student academic success in the Islamic Religious Education Study Program. This method was chosen because it produces a decision tree model that is easy to understand and has a high level of accuracy. The data used comes from student achievement indexes from semesters 1 to 5. The research results showed that the prediction system achieved 99.62% accuracy and achieved high recall precision across each class category. This demonstrates the effectiveness of the C4.5 algorithm in predicting student academic achievement and has the potential to serve as a valuable tool for decision-makers in higher education.

Crime, an unlawful act that contradicts ethics and norms, has now become a primary factor for the police in Lampung province. This presents a challenge for the police institution in predicting high crime rates. However, there are still many crimes that have not become the main focus of problem-solving at the Lampung Regional Police.This research aims to identify the types and criminal acts of crime with the highest recorded incidence in a crime dataset by performing classification using the Naïve Bayes algorithm. The data was obtained from investigators at the Directorate of General Criminal Investigation of the Lampung Regional Police, with a total of 12,034 JTP (Total Criminal Acts) and 7,518 PTP (Crime Resolution) data points for each type of crime, distributed across the Regional Police, City Police, and District Police throughout Lampung province. The classification process using the Naïve Bayes algorithm reveals the relationship between the work unit (Satker) and the type of crime handled, thereby identifying crime patterns based on the location where they are handled. The results of the research, which involved converting numerical data into binomial (binary) form using the "Numerical to Binominal" feature in Rapid miner, show that the analysis and modeling process, especially in algorithms like Naïve Bayes or decision trees, is more effective when using data in a binary format. Thus, the initial dataset can be visualized in the form of a , with the size of the text varying according to the level of each high-incidence crime; the larger the text, the more frequently or significantly the crime occurred or was reported. The application of this method can help in identifying patterns, dominant trends, and areas of focus for more targeted law enforcement efforts or crime prevention policies.

Credit scoring is an important element in decision-making for providing financing, especially for microfinance institutions. Several methods for predicting credit scoring include Decession Tree, Gradient Boosted, Neural Network, K-NN, and Rule Induction. This study aims to improve the accuracy of financing risk prediction by efficiently integrating historical data. The Neural Network (NN) algorithm is a machine learning algorithm consisting of neurons (nodes) connected to each other in several layers (input, hidden, and output). NN is used for pattern recognition, classification, regression, and complex non-linear modeling. The NN algorithm has the advantage of working well on large and diverse data and unstructured data. However, the NN algorithm has weaknesses such as overfitting and data dependence. In this study, the integration of the Sample Bootstrapping and Weighted Principal Component Analysis (PCA) methods is proposed to improve optimal accuracy in the NN algorithm. The Sample Bootstrapping method is used to reduce the amount of training data to be processed. The Weighted PCA method is used to reduce attributes. This study uses a financing customer dataset. The results of the study show that the integration of the NN algorithm with Sample Bootstrapping and Weighted PCA resulted in an accuracy increase of 1-3% (97%-99%) compared to other algorithms. Therefore, it can be concluded that the integration of the NN algorithm with Sample Bootstrapping and Weighted PCA produces better accuracy than other algorithms

Infectious diseases remain a dynamic and evolving public health threat, requiring data-driven approaches for early detection and targeted policy planning. This study aims to model spatio-temporal trends and clustering patterns of HIV transmission in Bogor Regency during the period 2020–2023 by utilizing a combination of unsupervised and supervised machine learning techniques. The dataset was obtained from the Bogor Regency Health Office and includes annual data on the number of HIV cases across 40 sub-districts. The research methodology consists of data preprocessing stages, clustering using the K-Means algorithm, and classification using a Decision Tree model. The preprocessing steps include data integration, attribute selection, temporal aggregation, handling of missing data, and normalization using Z-score. K-Means clustering is applied to identify hidden patterns in the development of HIV cases, resulting in three distinct clusters based on multi-year trends. The resulting cluster labels are then used as target classes in the supervised classification process. The Decision Tree classification model demonstrates high accuracy in predicting cluster membership, indicating a strong relationship between the temporal patterns of HIV cases and cluster identity. The integration of clustering and classification techniques provides a robust analytical framework for understanding the dynamics of HIV transmission, while also supporting the formulation of more precise, evidence-based, and region-specific public health interventions.

The growing interest in learning the Japanese language in Indonesia, driven by popular culture such as anime, creates a need to understand the effectiveness of different learning media. The non-uniform effectiveness of media for each individual poses a major challenge. Therefore, this study aims to analyze the effectiveness of both anime and textbooks by segmenting learner profiles and identifying key determinants of success using an artificial intelligence approach. This research employed a quantitative method through a questionnaire survey of 120 respondents. The data were analyzed in two stages: the K-Means Clustering algorithm was used to group respondents into learner profiles, and the Decision Tree algorithm was used to identify the most significant factors that differentiate these profiles. The analysis successfully identified three distinct learner profiles: "Intensive & Adaptive Learner," "Flexible Learner," and "Passive Learner." The decision tree revealed that the perception of textbook effectiveness and the frequency of anime use are the strongest predictors in determining a learner's profile, more so than theoretical learning style preferences. It is concluded that media effectiveness is highly dependent on the learner's behavioral and perceptual profile, which underscores the importance of a personalized approach in language education technology.

The Indonesia Pintar Program (PIP) is an educational assistance program for students from underprivileged families, but determining the eligibility of recipients still faces obstacles in the form of subjectivity and data imbalance. This study aims to classify the eligibility of high school students receiving PIP in Jambi City using data mining methods. The SMOTE technique was applied to overcome class imbalance, and Gain Ratio feature selection was used to determine important attributes. The dataset used consisted of 19,596 student data with a training data distribution of 70% and testing data of 30%. The classification process used the Naïve Bayes, Decision Tree (J48), and Random Forest algorithms with the Use Training Set, 5-Fold, and 10-Fold Cross Validation testing schemes. The results show that SMOTE improves model performance, but feature selection in some cases reduces accuracy. Overall, Random Forest without feature selection provides the best results with an accuracy of 93.33% and is recommended as the most effective model for objectively determining PIP recipient eligibility.

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.

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.

Selecting a faculty that aligns with students’ interests and talents is a strategic step in determining the success of higher education and future career paths. However, most vocational high school (SMK) students still face difficulties in identifying the most suitable faculty due to the lack of data-driven analysis. This study implements the C4.5 classification algorithm within data mining techniques to build an automatic and measurable faculty recommendation system. The dataset consists of attributes such as SMK major, interest level, aptitude test results, academic grade average, and gender, with the output being the recommended faculty. The C4.5 algorithm was chosen for its ability to generate a transparent and interpretable decision tree, which helps both guidance counselors and students understand the rationale behind the recommendations. The experimental results show that the constructed classification model achieved an accuracy rate of 88%, based on cross-validation testing using data from 12th-grade students. The implementation of this system is expected to serve as an objective tool in the faculty selection process and to promote a data-driven decision-making approach in secondary education environments.

Sleep is a fundamental human need that plays a crucial role in maintaining both physical and mental health. Poor sleep quality can trigger a variety of health problems, ranging from decreased concentration to an increased risk of chronic diseases. The complexity of factors influencing sleep quality—such as stress levels, heart rate, blood pressure, physical activity, and lifestyle—makes its assessment difficult through direct observation alone. Therefore, data mining approaches are increasingly utilized to identify relevant patterns in sleep-related data. This study aims to compare the performance of the C4.5 (Decision Tree) algorithm and the Naïve Bayes algorithm in predicting sleep quality using the Sleep Health and Lifestyle dataset, which contains information from 374 respondents. The research method applied is a quantitative comparative approach employing classification techniques with 10-fold cross-validation to ensure robust evaluation. Model performance is assessed using accuracy, precision, and recall metrics to provide a comprehensive understanding of the effectiveness of each algorithm. The findings indicate that the C4.5 algorithm achieves an accuracy of 96.26% and offers advantages in terms of interpretability through its decision tree visualization, enabling easier understanding of variable relationships. In contrast, the Naïve Bayes algorithm demonstrates superior predictive performance, achieving an accuracy of 98.66% along with consistently high precision and recall across nearly all classes. These results suggest that Naïve Bayes is more effective for predictive tasks involving sleep quality, while C4.5 remains highly valuable when the goal is to interpret variable interactions and decision rules. Overall, this research highlights the potential of data mining techniques in health informatics, particularly in improving the understanding and prediction of sleep quality, which in turn can contribute to better prevention and management of sleep-related health issues.