Recurrent Neural Networks (RNNs), particularly their Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) variants, are standard methods for modeling sequential data. However, their robustness is often limited when faced with non-stationary and heterogeneous time series data. This limitation is largely due to their reliance on symmetric loss functions such as mean squared error, which implicitly assume homogeneous data patterns. To address this, we propose a new framework, Expectile-based Recurrent Neural Network (E-RNN), which integrates expectile regression into the RNN architecture. We implement and compare two E-RNN variants, namely E-LSTM and E-GRU, to obtain the best forecast. . By leveraging the asymmetric least squares loss function, the E-RNN model is able to model various parts of the conditional data distribution, not just its central tendency. This allows forecasting across scenarios, ranging from pessimistic to optimistic, by adjusting the asymmetric parameter (τ), a value within the range (0, 1) where τ〈 0.5 yields pessimistic and τ〉 0.5 yields optimistic forecasts.. We demonstrate this methodology by forecasting Indonesia's quarterly economic growth data from 2001 to 2025. Empirical results show that the E-RNN model consistently exhibits superior performance, evidenced by lower Expectile-based Generalized Approximate Cross Validation (EGACV) scores for model selection and higher forecast accuracy. This superiority becomes particularly significant on more volatile quarter-to-quarter (qtq) data, highlighting the effectiveness of this framework in adapting to complex data dynamics and improving forecast reliability under uncertain conditions.

Traditional risk models often fail to capture extreme losses in interconnected global stock markets. This study introduces a novel approach, Expectile Regression Neural Network with Modified LASSO regularization (ERNN-mLASSO), to model nonlinear systemic risk. By analyzing five major stock indices (JKSE, GSPC, GDAXI, FTSE, N225), we identify distinct market roles: developed markets, such as the GSPC, act as risk spreaders, while emerging markets, like the JKSE, act as risk takers. Our network systemic risk index, SNRI, accurately captures systemic shocks during the COVID-19 crisis. More importantly, the model projects increasing global financial fragility through 2025, providing an early warning signal for policymakers and risk managers of potential future instability.

Accurate obesity risk classification is often hindered by the asymmetric and heteroscedastic nature of health data, where traditional mean-based machine learning models fail to capture critical distribution tails. This study addresses this gap by proposing a robust Expectile Random Forest (ERF) model, a novel ensemble architecture that integrates an expectile learning framework via the Asymmetric Least Squares (ALS) loss function for seven-level (multilevel) classification. Utilizing a dataset of 2111 empirical records, the sensitivity analysis identifies $\tau =0.7$ as the optimal configuration, achieving an overall Accuracy of 94.6 ± 0.7% and a Macro F1-Score of 94.5 ± 0.7%. This performance represents a significant quantitative improvement over state-of-the-art benchmarks, outperforming XGBoost by 1.8% and standard Random Forest by 3.9%. Feature importance analysis identifies body weight, age, and sedentary factors as primary predictors, while the ERF model demonstrates exceptional ordinal consistency and robustness against clinical outliers. These findings provide a superior methodological framework for developing precise medical decision support systems, shifting the paradigm from central-tendency predictions to tail-sensitive health risk mapping.

This study proposes a Reverse Mixed Data Sampling based on Expectile Regression (R-MIDAS-ER) framework for modeling Inflation at Risk and examining the role of economic growth in inflation risk dynamics. The proposed model addresses two central challenges in macroeconomic risk modeling: the mixed-frequency structure between monthly inflation and quarterly macroeconomic indicators, and the need to capture heterogeneous effects across the conditional inflation distribution. By integrating the Reverse MIDAS structure with asymmetric least squares expectile estimation, the model enables low-frequency real-sector information, particularly quarterly GDP growth, to explain high-frequency monthly inflation without imposing artificial aggregation or interpolation. Numerical experiments are conducted under Gaussian, heavy-tailed, skewed, contaminated, and heteroskedastic residual scenarios to evaluate finite-sample performance and robustness. The results show that R-MIDAS-ER provides stable recovery of the mixed-frequency transmission mechanism and achieves competitive or superior expectile forecasting performance, especially in upper-tail inflation risk regions. The empirical application to Indonesian data from January 2001 to June 2024 further demonstrates that inflation, GDP growth, financial stress, and financial conditions exhibit asymmetric and heavy-tailed behavior, supporting the use of distributional modeling. The R-MIDAS-ER model identifies several historically meaningful high-risk inflation episodes and reveals that GDP growth has a stronger association with upper-tail inflation risk than with lower-tail conditions. Financial conditions also exhibit asymmetric effects across expectile levels. Overall, the proposed framework offers an interpretable and distribution-sensitive approach for monitoring inflation vulnerability in mixed-frequency macroeconomic environments.

Indonesia’s transition toward digital tax administration has increased the importance of understanding how users experience public digital services. This study examines sentiment patterns and reported barriers in 820 Google Play reviews of the Coretax Mobile application posted during the January–April 2026 tax-filing period. Reviews were processed using a deterministic text-preprocessing procedure and analyzed through rating-derived sentiment classification, machine-learning models, and Latent Dirichlet Allocation topic modeling. Negative reviews dominated the dataset (93.9%), with a mean rating of 1.25 out of 5. Among the four classifiers, the Support Vector Machine achieved the highest macro-F1 score (0.4377), although its accuracy (0.9463) was only slightly higher than the majority-class baseline (0.9390), indicating limited performance under severe class imbalance. Topic modeling identified three recurring complaint themes: application malfunctions and recurrent failures (31.3%), account access and registration difficulties (36.2%), and tax-reporting and payment complexity (32.5%). These patterns suggest that user difficulties are concentrated not in a single stage of service delivery but across access, system reliability, and task completion. The findings support prioritizing user-centered redesign, technical reliability, and accessible user assistance, while remaining limited to self-selected Google Play reviewers rather than the broader population of Coretax users.

This study investigates the application of the Dynamic Time Warping (DTW) algorithm to cluster the ten stocks with the largest market capitalization on the Indonesia Stock Exchange as of February 2026. Unlike conventional distance metrics, DTW handles time lag nonlinearly to identify hidden temporal pattern similarities. Adjusted closing price data was obtained through web scraping from Yahoo Finance using the R programming language. The clustering procedure was performed using Ward's Hierarchical Agglomerative Clustering method, where the number of clusters was evaluated through Silhouette coefficient analysis and qualitative economic interpretability. While the highest Silhouette score suggested two clusters, a three-cluster solution was selected as the most representative structure to better capture the granular dynamics of different market sectors. The first cluster is dominated by the banking and energy sectors with stable growth trends. The second cluster includes cyclical industrial and infrastructure stocks with high volatility. The third cluster uniquely unites GOTO and UNVR stocks in a long-term bearish downward pattern, despite their origins in different sectors. These findings demonstrate that DTW is highly effective in uncovering cross-sector market dynamics, providing a more accurate basis for portfolio diversification strategies than traditional business sector classifications.