Credit card fraud detection remains challenging due to extreme class imbalance, evolving fraud patterns, and asymmetric misclassification costs. This paper presents a deployment-oriented evaluation and decision-calibration framework for transaction-level fraud detection on the public creditcard.csv dataset, assessed under a strictly chronological train–validation–test protocol that mirrors real-world operation. Our contribution lies in evaluation design and cost-aware decision calibration rather than architectural novelty, focusing on how probabilistic model outputs are translated into operational decisions under explicit cost constraints. A standard feed-forward deep neural network (DNN) is trained on the numerical features using a class-weighted binary cross-entropy loss, with early stopping guided by validation AUC–PR. At deployment time, the decision threshold is selected on the validation window by minimizing an empirical cost function that penalizes false negatives more than false positives. On the held-out test set, the proposed pipeline achieves a ROC-AUC of 0.9489 and a PR-AUC of 0.7813. We show that decision policy choice strongly affects operational outcomes: naive thresholding yields excessive false alarms, whereas validation-based cost-sensitive calibration substantially reduces expected loss. Under C_FN = 10 and C_FP = 1, the cost-optimal threshold yields an expected test cost of 190. Comparisons with logistic regression, random forest, and XGBoost under identical preprocessing, temporal splitting, and decision calibration show that tree-based ensembles remain highly competitive, while the evaluated DNN achieves comparable cost and precision–recall performance. Overall, the results highlight the importance of combining chronological evaluation with explicit cost-sensitive thresholding for practical fraud detection under severe class imbalance.

Indonesia continues to face digital inequality, both in terms of inter-provincial disparities and in comparison, toother countries in Southeast Asia, where it lags behind Vietnam and is far below Singapore. These disparities highlight spatialheterogeneity, where regional characteristics influence by Information and Communication Technology (ICT) capacitydifferently, and spatial dependence, where development in one province can spill over to others. Given the strategicimportance of ICT in driving sustainable growth, overcoming such inequality is crucial. This challenge is also stronglylinked to SDG 9 and SDG 10, which emphasize inclusive digital development as a pathway to reducing gaps across regions.This research seeks to examine the determinants affecting Information and Communication Technology Development Index(ICT-DI) across Indonesian province by contrasting traditional panel data models with spatial panel modeling techniques.Secondary data from Central Bureau of Statistics (Indonesia) for 2020-2023 were used. Descriptive analysis and thematicmapping were conducted, followed by estimation using the panel model, as well as spatial panel models including SpatialAutoregressive Fixed Effect (SAR-FE) and Spatial Error Model Fixed Effect (SEM-FE). The results indicate significantspatial dependence across provinces, confirming the relevance of spatial analysis. The SAR-FE model was identified asthe best model, explaining 98.47% of the variation in ICT-DI with the lowest MAPE value (1.1023). Population densitywas identified as the only significant positive factor, indicating that more densely populated regions tend to have betterICT infrastructure and capacity. The findings emphasize the novelty of applying spatial panel models to ICT analysis inIndonesia and underline their policy relevance. Considering dependence and heterogeneity enables policymakers to designmore inclusive and sustainable strategies, tailored to the unique priorities of each province, to reduce digital inequality.

This paper presents an innovative algorithm, termed the Fractional Optimization Algorithm (FOA), which utilizes the properties of fractional functions to enhance the integration of photovoltaic (PV) systems and distribution static compensators (D-STATCOMs) into distribution systems (DSs). The FOA employs a discrete-continuous encoding approach to determine the optimal placement and sizing of PV and D-STATCOM devices. A master-slave optimization framework is adopted, where the FOA operates in the master stage, and the successive approximations method is used for power flow analysis in the slave stage. The algorithm's efficacy is tested on 33- and 69-bus grids, demonstrating significant cost reductions over traditional optimization approaches such as the Vortex Search Algorithm (VSA) and the Sine-Cosine Algorithm (SCA). Furthermore, the FOA achieves superior computational efficiency, underscoring its promise as a robust optimization strategy.

The Poisson regression model is an important analytic tool that should be used in count data modeling. When the value of excess dispersion of variables, then the model is not appropriate to apply in case the value of the mean is not equal to the value of the variance of the Poisson distribution. The results are compatible with data when there is the use of Bell regression model. The number of zeros in the count data that is seen is very high. In this case, the Zero-Inflated Bell regression model is an alternative to the Bell regression model. Parameters of the Zero-Inflated Bell regression model are estimated mostly through the approach of maximum likelihood. In an extended linear model, in which the response variable is modeled by two or more explanatory variables, as in the Zero-Inflated Bell regression model, linear dependence is a threat in a real-life analysis. It reduced the maximum likelihood estimator in its effectiveness. In a bid to solve this issue, this paper explores the performance of the generalized shrinkage estimator in the zero-inflated Bell regression model. The superiority of the proposed approaches over the traditional maximum likelihood estimator is validated by results of the simulations and implementations.

Purpose: The Jakarta Provincial Government aims for 80% of MSMEs to be digitalized by 2025; however,empirical evidence on whether the intensity of Artificial Intelligence (AI) use truly enhances business performance in developing economies remains inconclusive. This study examines the relationship between AI usage intensity and business performance among MSMEs in Jakarta, while accounting for firm size variations. Method: A cross-sectional survey was conducted among 300 MSME owners/managers in Jakarta’s five administrative regions who had used at least one AI tool in the past 3 months. The Technology Acceptance Model (TAM) was extended with a Resource-Based View (RBV) perspective and firm size variables (micro, small, medium). Data were analyzed using CB-SEM with Maximum Likelihood estimation and FIML to handle missing data. Results: The model demonstrated an excellent fit (χ2/df = 1.13; CFI = 0.993; RMSEA= 0.021; SRMR = 0.018). However, AI usage intensity did not have a significant direct effect on business performance (β = 0.085; p = 0.113). Firm size had a substantial direct effect on performance (small: β = 0.446, p < 0.001; medium: β = 0.548, p < 0.001). Small firms tended to have higher AI usage intensity (β = 0.269, p < 0.001). Nevertheless, mediation analysis confirmed that AI usage did not function as a significant mechanism for improving performance among small or medium firms. Implications: The findings indicate the presence of adoption without impact—access to and intensity of AI use alone are insufficient; business value emerges only when complementary resources (dynamic capabilities, data governance, and human resource skills) are available. Policy programs should therefore integrate managerial training and infrastructure financing rather than merely providing technology license subsidies. Originality/Value: This study is among the earliest quantitative examinations in the ASEAN context exploring the relationship between AI usage intensity and performance among MSMEs, using a SEM–TAM approach that incorporates firm size as a contingency variable.

This study explores the use of Hidden Markov Models (HMM) combined with a hybrid ensemble method using boosting techniques for the classification of different types of electrocardiogram (ECG) signals, including Normal, Ventricular Tachycardia (VTach), Ventricular Fibrillation (VFib), and Bradycardia. The analysis focuses on leveraging Auto Correlation and Partial Autocorrelation (PACF) for feature extraction and enhancing the classification performance through a hybrid approach integrating HMMs and boosting. First, Wavelet-based filtering was applied to remove noise from the ECG signals, providing cleaner data for subsequent feature extraction. Both Autocorrelation (AC) and Partial Autocorrelation (PACF) were computed for the filtered signals. While AC provided general periodicity information, PACF offered a more precise analysis by isolating direct correlations at each lag, which was especially useful for differentiating irregular rhythms like VTach and VFib. We then implemented a hybrid ensemble method combining Hidden Markov Models (HMMs) with boosting techniques, such as AdaBoost and Gradient Boosting, to improve classification accuracy. The HMMs were used to model the sequential dependencies in the ECG signals, while the boosting algorithms were applied to optimize the performance of the ensemble by weighting and improving weaker classifiers iteratively. The study demonstrates that the combination of HMMs, boosting, and PACF provides a powerful and efficient method for automatic arrhythmia detection, achieving high precision and robustness across different ECG signal types.

The industrial fog computing setting is highly sensitive to challenges in efficiently scheduling tasks and offloading computing processes because of the heterogeneity of devices, their mobility, and fluctuations in resource availability. This work is motivated by the need to design a unified optimization mechanism for dynamic fog environments where both device and network heterogeneity severely affect task allocation. The proposed study aims to address the limitations of existing single-strategy offloading methods by developing a hybrid optimization framework that balances computational load, minimizes latency, and reduces energy consumption. The main contribution of this paper lies in introducing a hybrid GSA–EPO co-offloading model that achieves an adaptive trade-off between energy efficiency and service latency while maintaining scalability in large-scale fog computing systems. The hybridized framework integrates the Gravitational Search Algorithm (GSA) and Emperor Penguin Optimization (EPO) to reduce energy consumption, service latency, and penalties related to deadline violations. The model uses battery-aware willingness, memory constraints, and mobility-sensitive communication parameters to inform local execution, device-to-device offloading, and edge server offloading. The hybrid GSA–EPO algorithm combines the exploration capability of GSA with the convergence efficiency of EPO within a migration-based knowledge-exchange structure, thereby enhancing both convergence and diversity of solutions. Simulation results demonstrate that the proposed approach achieves up to 63% reduction in energy consumption and 75% reduction in delay compared to baseline methods, confirming the feasibility of hybrid metaheuristic strategies for reliable and efficient task co-offloading in dynamic fog computing networks.

Stunting is a growth impairment condition in children under five years old, resulting from chronic malnutrition and repeated infections, which causes them to be shorter than expected for their age. East Java is one of twelve priority provinces, with a stunting prevalence of 17.7% in 2023. Accurate identification of the factor influencing stunting is essential to support effective and targeted interventions. Given the spatial variability in these factors, conventional regression models such as Ordinary Least Squares (OLS) are inadequate. Geographically Weighted Regression (GWR) addresses this by allowing local variation, yet it assumes a uniform spatial scale across variables. This study employs the Multiscale Geographically Weighted Regression (MS-GWR) model, which enables each explanatory variable to operate at its own optimal spatial scale. The results show that MS-GWR with an adaptive Gaussian weighting function provides the best fit, with an AICc of 67.7426 and an R² is 0.79. Seven variable groups significantly influence stunting, including exclusive breastfeeding, early initiation of breastfeeding (EIB), and upper respiratory tract infections (URTIs), as well as combinations of these factors. These findings highlight the importance of formulating location-specific and context-sensitive policies that reflect the dominant characteristics of each region to effectively and sustainably accelerate stunting reduction.

This study develops and examines a comprehensive deep learning framework for the detection of multi-class healthcare fraud in National Health Insurance Scheme (NHIS) claims. We examined 20,388 NHIS healthcare claims revealing four specific fraud patterns: Phantom Billing, Wrong Diagnosis, Ghost Enrollee, and legitimate claims. Four different deep neural network architectures were developed and evaluated: Simple NN, Deep Wide NN, Regularized NN, and Residual NN, in addition to ensemble methods. The Simple Neural Network achieved the highest overall performance, with a test accuracy of 79.84% and an F1-macro score of 77.76%. Despite possessing only 100,324 parameters (five times fewer than the Wide Deep Neural Network), it outperformed more complex designs while achieving the fastest training time of 40.61 seconds. Multiclass analysis demonstrated exceptional performance in Ghost Enrollee detection (97.84% F1-score) and moderate performance in Phantom Billing detection (61.15% F1-score).

We introduce the notion of total distance irregular labeling, called the local total distance irregular labeling. All edges and vertices are labeled with positive integers 1 to k such that the weight calculated at the vertices induces a vertex coloring if two adjacent vertices has different weight. The weight of a vertex $u\in V(G)$ is defined as the sum of the labels of all vertices adjacent and edges incident to $u$ (distance $1$ from $u$). The minimum cardinality of the largest label over all such irregular assignment is called the local total distance irregularity strength, denoted by $tdis_l(G)$. In this paper, we established the lower bound of the local total distance irregularity strength of graphs $G$ and determined exact values of some classes of graphs namely path, cycle, star, bipartite complete, fan and sun graph.