Publication Search

Private vehicles are a frequently used mode of transportation because they are considered more practical. However, using private vehicles carries several risks, such as traffic accidents due to drivers losing focus on the road due to other activities, such as making calls on smartphones, drinking, or operating the radio. Approximately 90% of accidents are caused by human error. Convolutional Neural Network (CNN) is a type of neural network commonly used on image data. CNN is often used for image classification due to its high performance and accuracy. Therefore, this study aims to analyze the performance of CNN for the classification of distracted driving activities. The results show that the CNN model is able to effectively classify images of distracted driving activities, with an accuracy of approximately 99% across all datasets and across all input image size variations. Furthermore, the results of this study also show that differences in right-hand and left-hand drive datasets do not significantly affect model accuracy. Variations in input image size also do not significantly affect model accuracy, but do affect the training duration.

Real time decision making applications, such as those used in autonomous vehicles, smart cities, and industrial IoT, require fast, scalable, and accurate analytics to ensure timely responses and optimized operations. Traditional cloud-based systems face significant challenges in meeting these requirements due to high latency, limited scalability, and bottlenecks in data processing. This study explores the use of a hybrid Edge Cloud architecture to optimize End to end machine learning (ML) pipelines for real time applications. The proposed system offloads time-sensitive tasks to edge devices, while computationally intensive processes are handled by the cloud, ensuring efficient use of resources and reduced latency. Experimental results demonstrate that the hybrid model reduces inference latency by up to 70% compared to cloud-only systems, while maintaining model accuracy and increasing throughput. Additionally, the scalability of the hybrid architecture is highlighted, as it can handle large-scale data streams and adapt to varying workloads. The findings show that hybrid Edge Cloud architectures are well-suited for applications where fast decision making is critical, such as autonomous systems and real time analytics in smart cities. However, challenges remain in managing resources across edge and cloud systems, particularly in balancing computational loads and ensuring system reliability. Future research should focus on optimizing task partitioning, integrating advanced edge AI models, and exploring the use of 5G networks to enhance performance further. Overall, the study demonstrates the potential of hybrid Edge Cloud systems in overcoming the limitations of traditional cloud-based ML pipelines and provides insights into the future of real time data processing.

This study aims to analyze and project consumer prices of cabbage commodities at four levels: Ngawi Regency, Pacitan Regency, East Java Province, and nationally, using the additive Holt–Winters forecasting model. Monthly price data for the period January 2020–December 2024 were used to capture the dynamics of levels, trends, and seasonal patterns that affect price fluctuations. Model performance was evaluated using the Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) indicators. The results showed differences in model accuracy between regions. East Java Province produced the best performance with the lowest MAE and RMSE values, indicating a more stable price pattern that was easier for the model to capture. In contrast, Ngawi Regency showed the highest volatility, resulting in greater forecasting errors. Pacitan Regency displayed a relatively consistent seasonal pattern with moderate accuracy, while national data showed smoother fluctuations due to the aggregation effect. Overall, the additive Holt–Winters model is effective for short-term projections in regions with low to moderate variability, but is less optimal in regions with highly volatile price dynamics.

Information system security faces serious challenges due to increasingly complex cyber attacks. Intrusion Detection Systems (IDS) require efficient approaches to handle high-dimensional data such as the NSL-KDD dataset with 41 features. This study aims to implement the Genetic Algorithm (GA) for feature selection on the NSL-KDD dataset to improve the efficiency and accuracy of network attack detection. The method used is computational experimental research, involving data preprocessing, GA implementation for feature selection, building a classification model using Random Forest, and performance evaluation based on accuracy, precision, recall, F1-score, and computation time. The results show that GA successfully reduced features from 41 to 12 features (70.7% reduction), significantly improving computational efficiency. However, model accuracy slightly decreased from 0.4973 to 0.4951, indicating that while GA is effective for feature selection, the elimination of certain features may reduce classification capability. The implication of this study is that GA can be used as a tool to simplify intrusion detection models, but it should be combined with parameter optimization and data imbalance handling to achieve more optimal performance.  

Effective management of fixed assets plays a crucial role in maintaining the reliability and transparency of a company’s financial reporting. Errors in the capitalization process can lead to misstatements in financial statements and affect investment decisions. This study aims to analyze and forecast asset capitalization trends using the Autoregressive Integrated Moving Average (ARIMA) model. The research utilizes monthly recap data of asset capitalization recorded during the Settlement to Fixed Asset process from January 2021 to August 2025. The data were processed through several stages, including stationarity testing, model identification, parameter estimation, and model accuracy evaluation. The findings indicate that the data are stationary without differencing (d = 0). From several candidate models, ARIMA(0,0,3) was identified as the best model based on the lowest AIC value of 39.76. The selected model was then applied to predict asset capitalization values for the next ten periods, resulting in forecasts ranging from 1.12 to 1.56 trillion rupiah. Model evaluation showed a MAPE of 29.01%, which implies a moderate forecasting accuracy. Consequently, the ARIMA model can be considered a suitable analytical tool for monitoring and forecasting asset capitalization quantitatively.

This study aims to examine the influence of transformational leadership, work motivation, and organizational culture on organizational commitment at PT Kembang Jawa Permai. The research employs a quantitative approach with an associative causal design, using a survey method and data collected through a questionnaire distributed to all 46 employees as respondents. The sampling technique used is saturated sampling, considering the small population size. Data analysis was performed using multiple linear regression analysis with validity, reliability, and classical assumption tests (normality, multicollinearity, heteroscedasticity, and linearity) conducted beforehand to ensure model accuracy. The results indicate that transformational leadership, work motivation, and organizational culture each have a positive and significant effect on organizational commitment, both partially and simultaneously. Among the three variables, organizational culture shows the most dominant influence, indicating that strong organizational values and teamwork orientation play a key role in strengthening employee commitment. The coefficient of determination (R²) of 0.848 implies that 84.8% of the variation in organizational commitment can be explained by the three independent variables. This study contributes to human resource management theory and provides practical insights for organizations to enhance employee commitment through effective leadership, motivation, and cultural reinforcement.

Human Resource Management (HRM) plays a strategic role in improving organizational competitiveness through proper management of employee placement, training, and performance evaluation. To support the achievement of these goals, a predictive model is needed that can provide an accurate picture of employee performance. This study utilizes a Human Resource Management (HRM) dataset of 1,200 data and applies several classification algorithms to compare their effectiveness, namely J48 or C4.5, Random Forest, Naive Bayes, K-Nearest Neighbor (KNN), Logistic Regression, and Support Vector Machine (SVM). To obtain more optimal results, this study uses resampling techniques and attribute selection methods with a correlation attribute eval approach, so that class distribution can be more balanced and model accuracy increases. From the test results, the Decision Tree J48 algorithm showed the best performance with an accuracy level reaching 95.41%, a kappa value of 0.8925, a mean absolute error (MAE) of 0.0432, a precision of 0.955, a recall of 0.954, and an area under the ROC curve of 0.964. These findings indicate that J48 has excellent predictive capabilities compared to other algorithms. Furthermore, this study also found that the most influential variables in determining employee performance include the percentage of the last salary increase (EmpLast Salary Hike Percent), the level of work environment satisfaction (Emp Environment Satisfaction), the length of time since the last promotion (Years Since Last Promotion), and experience in the current role (Experience Years in Current Role). Overall, the results of the study indicate that the C4.5 algorithm with the application of the resampling technique can be an optimal solution in building an employee performance prediction system. Thus, this model has the potential to be a strong basis for managerial decision-making, particularly in designing HR development strategies and policies to improve organizational performance.

This study explores the integration of blockchain technology with federated learning (FL) to enhance cross-organizational healthcare analytics while ensuring privacy and data security. Federated learning allows multiple institutions to collaboratively train machine learning models without sharing sensitive patient data. Instead, local data is used to train models, and only model parameters are exchanged. However, privacy concerns and data sharing inefficiencies have hindered broader healthcare collaboration. Blockchain, a decentralized ledger technology, addresses these concerns by ensuring data integrity and transparency, providing an immutable and tamper-proof record of all transactions. This study investigates how the combination of blockchain and federated learning can overcome these challenges, facilitating secure and efficient data sharing between healthcare institutions. The study uses synthetic multi-institution healthcare datasets to simulate real-world collaboration scenarios. The blockchain-enabled federated learning system ensures that no raw patient data is shared, significantly reducing the risk of privacy breaches while still allowing healthcare institutions to collaborate on predictive model development. The results show that while there is a slight decrease in model accuracy compared to centralized methods, the trade-off is outweighed by the privacy and security benefits. Blockchain’s integration ensures that model updates are transparent, enhancing trust between institutions and reducing concerns about data integrity. Moreover, the use of blockchain’s smart contracts automates and enforces compliance, further streamlining collaboration. This research contributes to the field by demonstrating how blockchain-integrated federated learning can create a secure, scalable, and privacy-preserving framework for collaborative healthcare analytics. The findings underscore the potential for this approach to enhance healthcare outcomes and improve decision-making across institutions while ensuring patient data protection.

This study explores the development and application of a digital twin-driven cybersecurity risk assessment model for Industrial Internet of Things (IIoT) networks. The increasing complexity and interconnectivity of IIoT systems have expanded the attack surface, making them vulnerable to a wide range of cyber threats. The digital twin model addresses this challenge by creating real-time virtual replicas of physical systems, which can simulate and predict network vulnerabilities and attack vectors. The model uses machine learning algorithms and real-time data to simulate cyberattacks, including Distributed Denial of Service (DDoS), malware, and data breaches. By providing continuous monitoring and dynamic risk predictions, the digital twin model enhances the resilience of IIoT networks compared to traditional cybersecurity frameworks. The findings indicate that the model's ability to predict potential cyber threats and simulate various attack scenarios provides a more proactive and accurate approach to cybersecurity in IIoT environments. Additionally, the study highlights key mitigation strategies, including adaptive security mechanisms, real-time anomaly detection, and the use of lightweight encryption for resource-constrained devices. Despite its effectiveness, challenges such as computational requirements, integration with legacy systems, and scalability were identified. This research underscores the strategic importance of digital twin models in securing IIoT systems and advancing Manufacturing 4.0 ecosystems. Future research should focus on enhancing model accuracy, expanding its application to diverse industrial sectors, and improving interoperability with legacy systems to further strengthen the security posture of IIoT networks.

Fraud in digital transactions has become a serious issue threatening the security and integrity of the fintech and e-commerce sectors. To address this problem, machine learning technology has emerged as an effective solution for automatically detecting anomalies and fraudulent transactions. This study aims to analyze the application of machine learning algorithms, specifically Support Vector Machine (SVM), Random Forest, and Ensemble Learning, in detecting fraud in digital transactions. The research adopts a quantitative approach with experimentation, testing the effectiveness of the three algorithms using a digital transaction dataset consisting of both fraudulent and non-fraudulent transactions. The results show that the Random Forest algorithm performs the best in terms of accuracy and recall, followed by Ensemble Learning, which enhances fraud detection performance by combining multiple prediction models. Meanwhile, SVM demonstrates satisfactory performance but is prone to overfitting issues when handling large and complex datasets. The study also finds that the problem of imbalanced data can affect model accuracy, and data balancing techniques such as oversampling are required to improve fraud detection performance. Overall, the findings suggest that machine learning, particularly Random Forest and Ensemble Learning algorithms, can be relied upon to improve fraud detection in digital transactions. However, challenges such as model interpretability and the need for periodic algorithm updates still need to be addressed to enhance the effectiveness of fraud prevention systems in countering the ever-evolving nature of fraud.

Indonesia is committed to achieving zero hunger as one of the goals of fulfilling the Sustainable Development Goals (SDGs) where this commitment focuses on addressing the problem of food availability but also ensuring that every individual has access to sufficient, nutritious, and safe food throughout the year for everyone. However, reviewing the current conditions in Indonesia, there is still an imbalance in food availability that will cause food vulnerability. Therefore, a prediction of food vulnerability in the future is needed where discriminant analysis is one of the appropriate statistical methods to analyze qualitative dependent and quantitative independent variables. This study uses secondary data from the official website of the food agency and the central statistics agency. The results of the study show that the characteristics of the data have small variations, asymmetric distribution, and there are outliers in several categories. The assumptions of multivariate normality, the suitability of the dependent variables, and the identity of the variance-covariance matrix have been met. Through discriminant analysis, the variables of the percentage of poverty and the percentage of households with access to clean drinking water are proven to significantly affect the IKP category. The discriminant model produces one significant function that is able to group the IKP category with a model accuracy rate of 86.8% and a classification accuracy of 64.7%.

The development of culinary MSMEs in Banjarsari District, Surakarta City has developed rapidly, but the increase in the number of culinary MSMEs in Banjarsari District, Surakarta City is not necessarily followed by an increase in the financial performance of these MSMEs  This study aims to find out and analyze the influence of financial literacy, financial technology, entrepreneurial orientation, and innovation towards the financial performance of culinary MSMEs in Banjarsari District, Surakarta City. Data collection in this study uses a questionnaire distributed to respondents. The sample in this study amounted to 95 culinary MSMEs in Banjarsari District, Surakarta City with the type of sampling, namely purposive sampling with the consideration that the MSMEs have been running for at least 2 (two) years. The analysis methods used in this study are descriptive analysis, multiple linear regression analysis, t-test, F test (model accuracy test), and determination coefficient test (R2).  The results showed that the determination coefficient (adjusted R Square) is 0.548. Means This means that the amount of contribution of the influence of the independent variable X1 (financial literacy), X2 (financial technology), X3 (entrepreneurial orientation) and X4 (product innovation) towards Y (performance finance) by 54.8%. The rest (100% - 54.8%) = 45.2% is influenced by other variables outside the model such as the work environment, company size, market competitiveness, operational costs, working capital, etc.

This research focuses on using the K-Nearest Neighbor (KNN) algorithm to model student satisfaction with campus services. The study finds that the quality of the dataset strongly influences the accuracy of the KNN classification results. Factors such as data cleanliness, balanced class distribution, and sufficient training data volume are highlighted as crucial for a successful model. The research also emphasizes the significance of proper feature selection in enhancing classification performance, suggesting that irrelevant features can introduce noise and decrease model accuracy. The model was evaluated using a dataset of 1032 data points and K=5, achieving an accuracy of 93.72%. While the model performed well for certain classes such as "Very Good" and "None", challenges were encountered in classifying the "Fair" and "Deficient" classes. The study concludes that KNN is effective in identifying student satisfaction patterns but highlights the need for improvements in accurately classifying these challenging classes. Ultimately, the research underscores the importance of data quality and feature selection in enhancing the performance of classification models for student satisfaction analysis.

This research aims to determine the influence of stock prices, stock returns, and capital market training on interest in investing in shares in the capital market among students at the Faculty of Economics, Slamet Riyadi Surakarta University. The population of this research were students from the Faculty of Economics, Slamet Riyadi Surakarta University, from whom a sample of 95 respondents was taken using a purposive sampling method with the criteria being that students had/are currently taking capital markets courses. The analytical methods for this research are descriptive analysis, multiple linear regression analysis, t test, F test (model accuracy test), and coefficient of determination test. The results of the research prove that stock prices, stock returns and capital market training partially have a positive and significant effect on interest in investing in shares in the capital market among students at the Faculty of Economics, Slamet Riyadi Surakarta University.

This study aims to apply the K-Nearest Neighbors (KNN) algorithm to predict stunting status in young children based on height and weight data. Stunting is a growth failure condition caused by chronic malnutrition that negatively impacts children's physical and mental development. The dataset includes height, weight, and stunting status of children. The results show that the KNN model with k=3 achieved 100% accuracy on the test data. Evaluation using the confusion matrix and classification report indicates perfect precision, recall, and F1-score for each class. Data normalization with StandardScaler improved the model's performance by ensuring all features are on the same scale. The KNN algorithm proves to be a simple yet effective method for predicting stunting, demonstrating significant potential for early detection and health intervention in children. This study recommends using a larger and more diverse dataset, as well as incorporating additional relevant features to enhance model accuracy. Implementing the model in a web or mobile application is also suggested to assist healthcare professionals in the field.

Big data has revolutionized the landscape of natural sciences by providing extensive datasets that enable deeper insights and more accurate predictions. However, effectively analyzing such vast and complex data requires optimized machine learning algorithms tailored to specific applications. This study focuses on enhancing the performance of machine learning models in big data analysis for applications in natural sciences. The research aims to identify key optimization techniques, including feature selection, hyperparameter tuning, and algorithm customization, to improve model accuracy and computational efficiency. A combination of supervised and unsupervised learning approaches was applied to real-world datasets in fields such as climate science, genomics, and ecology. The findings demonstrate significant improvements in predictive accuracy and processing speed, highlighting the potential of optimized machine learning techniques in solving complex problems in natural sciences. The implications of this research extend to more efficient resource utilization and improved decision-making in scientific exploration and environmental management.

This research aims to develop and apply a Convolutional Neural Network (CNN) algorithm to detect handwritten Roman numerals. Handwriting recognition is a classic challenge in the fields of image processing and machine learning, especially for less common characters such as Roman numerals. In this research, we use data augmentation techniques to increase the diversity and number of datasets used in model training, which is expected to increase model accuracy and generalization. The dataset used consists of 1,120 images for testing and 280 images for validation, each of which is divided into 14 classes of Roman numerals I, II, III, IV, V, VI, VII, VIII, IX, X, L, C, D , and M. Image data was created directly using simple software, namely Paint version 6.3. This research uses the Python programming language and Google Colab as a computing platform. Model training was carried out for 300 epochs and showed significant accuracy in the 150th to 300th iteration. The results at the 300th epoch show an accuracy of 0.9607 and a loss of 0.1162. The implementation of this algorithm shows significant potential in practical applications, such as in the fields of education and historical documentation. The conclusion of this research is that data augmentation is an effective technique to improve the performance of CNN models in detecting handwritten Roman numerals.

The advent of the Internet as an effective means for resource sharing has consequently, led to proliferation of adversaries, with unauthorized access to network resources. Adversaries achieved fraudulent activities via carefully crafted attacks of large magnitude targeted at personal gains and rewards. With the cost of over $1.3Trillion lost globally to financial crimes and the rise in such fraudulent activities vis the use of credit-cards, financial institutions and major stakeholders must begin to explore and exploit better and improved means to secure client data and funds. Banks and financial services must harness the creative mode rendered by machine learning schemes to help effectively manage such fraud attacks and threats. We propose HyGAMoNNE – a hybrid modular genetic algorithm trained neural network ensemble to detect fraud activities. The hybrid, equipped with knowledge to altruistically detect fraud on credit card transactions. Results show that the hybrid effectively differentiates, the benign class attacks/threats from genuine credit card transaction(s) with model accuracy of 92%.