In the present work, a statistical modeling of the % increase in von-Mises and Tresca stress was done for a structural steel pipe containing a quarter-ellipsoidal internal corrosion defect subjected to internal hydrostatic pressure. An analytical solution was developed considering the internal corrosion defect length (z), width (x), depth (y), and internal hydrostatic pressure (P) as independent variables. The finite element analysis (FEA) simulated dataset was used for modeling. The analytical equations demonstrated high predictive accuracy, with coefficients of determination (R²) of 98.64% for von Mises stress and 98.38% for Tresca stress. The stress distribution from FEA was nearly similar quantitatively for both stresses; however, their patterns exhibited noticeable differences in their respective profiles inside and outside of the curvature of the internal corrosion defect. The contour plots for the percentage increase in von-Mises and Tresca stresses revealed both elastic and plastic regimes. The maximum percentage increase in von Mises stress remained slightly below the ultimate tensile strength of the specimen. In contrast, the contour plots for Tresca stress indicated a limited region in which the stress exceeded the ultimate tensile strength. For the linear effect, the geometrical variable x (width of the internal corrosion defect) exhibited a negative T-value, indicating that an increase in x reduces the percentage increase in both von Mises and Tresca stresses. The significance of standardized effects for von-Mises stress was in the order of P > y > z > x > zy> yp> zz> xy> xP> zP> xx. However, for the Tresca stress, the order was P > y > x > z > zy> yp> xy> zz> xP> zP> xx. The steel pipe with an internal corrosion defect of minimum length (z = 30 mm), maximum width (x = 134 mm), and minimum depth (y = 0.5 mm) exhibited the lowest von Mises and Tresca stresses. However, the maximum stresses were observed for a defect with maximum length (z = 454 mm), minimum width (x = 26 mm), and maximum depth (y = 1.5 mm).

Accurate Remaining Useful Life (RUL) prediction is essential for reliable battery management and cost-effective maintenance in lithium-ion energy storage applications. This study proposes a leakage-safe learning framework for battery RUL prediction from numerical cycling features, emphasizing realistic generalization to unseen batteries through group-aware splitting. The core contribution is a proper out-of-fold (OOF) stacking ensemble trained under GroupKFold to prevent information leakage across samples from the same battery while enabling an effective meta-learner to combine diverse learners. Experiments are conducted on a public Battery RUL dataset from Kaggle, and performance is evaluated using complementary regression metrics (R², MAE, RMSE) and robust percentage errors (sMAPE, WAPE). Results show that the proposed stacking approach (OOF GroupKFold with Meta-XGBoost) achieves the best overall performance (R² = 0.999550, MAE = 5.297907, RMSE = 6.834146, sMAPE = 3.080569, WAPE = 0.958618), outperforming strong baselines including XGBoost and Random Forest. These findings confirm that leakage-safe group-aware stacking can significantly enhance accuracy and stability for battery RUL prediction in practical deployment settings.

Forecasting train passenger demand is essential for supporting strategic decision-making and optimizing resource allocation in the transportation industry. This study aimed to develop a predictive model for the number of passengers on the Surabaya-Jakarta train route using the Extreme Gradient Boosting (XGBoost) algorithm. Owing to the non-linear nature of the historical count time series data (January 2019 to December 2024) and the significant disruptive impact of the COVID-19 pandemic, traditional linear models such as ARIMA were considered less appropriate. To optimize the XGBoost model, we comparatively evaluated two distinct input approaches: significant Partial Autocorrelation Function (PACF) lag and the sliding window method. Hyperparameter tuning was conducted via grid search, and the models were rigorously evaluated using Time Series Cross-Validation to prevent information leakage. Furthermore, the study compared recursive and direct multi-step forecasting strategies to project passenger volumes for the next 12 months. The analysis revealed that the sliding window approach with a window size of 4 yielded the best performance on the testing data, achieving a Mean Absolute Percentage Error (MAPE) of 10.94% and significantly outperforming the PACF lag method, which was prone to overfitting. Additionally, recursive forecasting is more rational and effective at capturing complex seasonal patterns and short-term fluctuations than direct forecasting. The final 12-month projection for 2025 indicates clear seasonal fluctuations, with a low in March (10,319 passengers) and a peak in November (20,932 passengers), providing a data-driven foundation for the train company to proactively optimize capacity planning, operational scheduling, and human resource management in the future.

The effectiveness and efficiency of a machine learning model can be improved by feature selection, especially for high-dimensional datasets such as in cybersecurity. The proposed approach utilizes an enhanced version of the Rainbow agent with a memory storage structure. The suggested approach is assessed using two benchmark datasets namely RT-IoT2022 which is targeted towards IoT network security and the Android Malware Detection dataset which is meant for mobile security. The specification of the reinforcement learning model has been trained for 20 epochs and it is progressively enhanced through feature subsets to enhance classification accuracy. The results show that the AUC scores continuously increase were the one for RT-IoT2022 achieves 0.91 and Android at 0.93. Three well-known classifiers XGBoost, Random Forest and multi-layer perceptron (MLP) are used to test the power of the selected features. The outcome evaluation on RT-IoT2022 dataset shows that Random Forest achieved maximum accuracy (99.48%), followed by XGBoost (99.16%), while MLP secured 94.04% accuracy. In the Android malware dataset, XGBoost model gave the best accuracy of 89.50%, followed closely by Random Forest with 87.00% and MLP with 86.50%. This clearly shows that reinforcement learning based feature selection enhances accuracy and reduces computation. The research emphasizes utilizing dynamic feature selection in any cyber security application. The future will experiment with incorporating deep reinforcement learning as well as hybrid selection.

In this paper, a hybrid multi-layer machine learning model is proposed for early prediction of PD using voice data. The suggested technique combines a number of base learners namely Random Forest, XGBoost, LightGBM, and CatBoost through ensemble methods like Voting Classifier, and Stacking Classifier. A thorough feature engineering, from correlation analysis to feature selection and polynomial feature extraction, was applied to improve the model’s quality. Experiments results show that the hybrid model obtained accuracy of 91.30% with high precision and recall especially for Class-1(Parkinson’s disease). Although very strong in its entirety the limited ability to discriminate healthy subjects (Class 0) (recall around 0.75). Among the selected models, Voting Classifier and CatBoost were robust and presented generalization well for all performance measures. The contribution of this work is two-folds: it demonstrates that ensemble models can be used to enhance diagnostic accuracy, and it addresses three main issues inherent in Parkinson's disease detection. Optimization of the performance of the model in discriminating healthy from Parkinson’s disease individuals is highly valuable, with an emphasis on the generalization and clinical usage.