Non-Technical Losses (NTLs) of electric energy compromise the operational efficiency and sustainability of the electrical system, particularly in the residential sector. This study addresses this problem by developing a predictive model that can estimate energy consumption and detect anomalous patterns. For this purpose, data were collected from the Plataforma Nacional de Datos Abiertos and the Osinergmin website. The study integrates two approaches: ARIMA, which is used to represent time series with well-defined seasonal patterns, and an approach based on XGBoost to represent non-linear behavior in more heterogeneous consumption intervals. The results suggest that ARIMA demonstrated optimal performance in stable cases, with errors close to zero in several cases, where the most representative systems are SR0148 with Mean Absolute Error (MAE) = 0.000124 and Root Mean Square Error (RMSE) = 0.003549, and SE1095 with MAE = 0.000287 and RMSE = 0.004481. XGBoost, on the other hand, reached its maximum performance in the interval "From 1 to 30 kWh", with MAE = 2.81, RMSE = 5.80, and a Coefficient of Determination (R²) of 0.13. This validates the effectiveness of the proposed approach based on the integration of more than one algorithm to identify electric consumption anomalies.

Indonesia is a strategically significant archipelagic nation with approximately two-thirds of its territory consisting of water. This geographical condition gives Indonesia greater potential than other countries. Amphibious aircraft serve as an alternative solution for the mobility and utility of residents living in remote areas surrounded by water, ensuring that these individuals can benefit from fair and equitable government services and their continuous development is necessary to maximize their functionality. This development encompasses various aspects, including the accuracy of flight trajectory estimation. Several machine learning methods have been developed for estimating the flight trajectory of amphibious aircraft. The present study implements and compares two filtering methods, namely the Ensemble Kalman Filter (EnKF) and the Extended Kalman Filter (EKF). The simulation result indicates that the EnKF method achieved a Root Mean Square Error (RMSE) value of 0.0214 for estimating the drag coefficient (CD) and the EKF method attained 0.0186. Furthermore, the EnKF method recorded an RMSE value of 0.0015 for estimating the yaw coefficient (CY), while the EKF method achieved 0.0012.

Cyberattacks using bad URLs are on the rise and are quickly becoming a key means of disseminating contemporary online dangers. Numerous cybersecurity agencies have reported a notable increase in Advanced Persistent Threat (APT), malware, and phishing activities that leverage URLs to spread their attacks. This circumstance emphasizes the importance of creating effective and flexible detection systems to thwart increasingly intricate attack patterns. A previous study proposed a malicious URL detection algorithm based on neural networks that achieved 97% accuracy. However, this model's reliability against dynamic attacks is still limited because it has not been verified using actual network data and still has a high false prediction detection rate (18.13%). To address these limitations, this study proposes OFERCE (Optimized Rule-based Detection for Malicious URLs), an optimized rule-based feature extraction framework that integrates adaptive feature selection based on mutual information, data-balancing strategies (SMOTE and class weighting), and comprehensive lexical rule-based features. Additionally, OFERCE incorporates hyperparameter tuning to ensure that the underlying machine-learning models operate at their optimal configuration, enhancing generalization capability and reducing overfitting during real-world evaluation. According to experimental results, OFERCE improves the performance of the Logistic Regression model to 99% accuracy and reduces the average error detection rate by up to 30%, demonstrating consistent and reliable performance across multiple categories of URL-based threats.

To ensure safety, autonomy, and context-aware control, reliable state recognition is still a challenge for stair-climbing wheelchairs, which offer mobility in environments with steps, curbs, and landings. This study uses instantaneous Inertial Measurement Unit (IMU) data to develop a simplified, window-free method for classifying wheelchair stair-related activities. Instead of sliding-window preprocessing or temporal sequence modeling, this study uses an 18-channel feature set that includes orientation, gyroscope, accelerometer, magnetometer, linear acceleration, and gravity signals. Stratified evaluation and several metrics, such as accuracy, macro-F1, Matthews Correlation Coefficient (MCC), ROC-AUC, and per-class precision-recall, were used to systematically benchmark eight classifiers: Multinomial Logistic Regression (MLR), Gaussian Naïve Bayes (GNB), K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), RBF-kernel Support Vector Classifier (SVC), a compact Multilayer Perceptron (MLP), and an adaptive CatBoost configuration. The results show that CatBoost performed almost flawlessly (Accuracy = 0.999, MCC = 0.999), closely followed by compact MLP. RF, KNN, and SVC formed a solid middle tier. In a window-free regime, feature importance analysis showed that instantaneous gyroscope and linear acceleration made very little contribution, while orientation and magnetometer channels were found to be the most crucial features. These results show that accurate and computationally efficient recognition of stair-climbing wheelchair states is possible by features driven by posture and heading. The proposed method facilitates low-latency embedded deployment and identifies areas for future development, such as lightweight temporal enhancements, angle encoding, and magnetometer calibration.

Intrusion Detection Systems (IDS) play a vital role in securing Internet of Things (IoT) networks against cyber-attacks. Previous work used a binary classification approach to detect only Denial-of-Service (DoS) attacks. This paper presents an improved multiclass IDS that identifies multiple attack types. The proposed approach uses simulated IoTID20-based traffic datasets to classify Normal, Denial-of-Service, Mirai botnet, Man-in-the-Middle (MITM), and Scan activity attacks. The proposed approach uses a combination of feature selection methods, including correlation-based filtering and a genetic algorithm, followed by machine learning classifiers such as Decision Tree (DT), Random Forest (RF), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM). Experimental results show a significant improvement in classification performance. The RF classifier with Genetic Algorithm (GA)-based feature selection achieved the highest accuracy (96.5%), followed closely by DT and SVM. Therefore, based on our theoretical and experimental comparisons, the proposed approach could be a practical step toward deploying more robust, realistic IDS models in IoT environments.

The need for resilient and adaptable training solutions in the aviation industry requires innovation beyond traditional methods. Virtual Reality (VR) provides immersive simulation, while Artificial Intelligence (AI) enables greater personalization and interaction. This paper presents the development of an evolving self-designed VR platform for aviation training, outlining a scalable design process and its progression from baseline applications to AI-augmented functionality using commercial aircraft (Airbus A320) case studies. The proposed method establishes the principles for the development of the core VR platform to ensure scalability, applicability, and efficiency across related projects. Progressive development is illustrated through case studies ranging from VR-based landing-gear maintenance to VR/AI-integrated aircraft walkaround procedures. Early feedback indicates positive perceptions of the platform among trainees, even within short pilot interventions. The combination of VR environments and AI-powered extensions, supported by an extensible architecture, offers a promising path toward more customized, effective, and intelligent training, with the potential for operational efficiency for aviation stakeholders.

With the increasing risk of cyberattacks, cybersecurity has now become a major area of the Internet of Things (IoT). Cyberattacks on IoT devices can result in serious concerns, including unauthorized access, disruption of critical services, and data breaches. IoT security aims to decrease the risk for enterprises and users by ensuring the safety of IoT resources and user privacy. Novel cybersecurity tools and technologies have enabled significant improvements in IoT security management. Recently, Artificial Intelligence (AI) has been applied to identify cyber threats, rapidly analyze millions of events, and detect multiple threats. Deep Learning (DL) has proven effective and offers several advantages for addressing IoT cybersecurity challenges. This study proposes an Optimized Dimensionality Reduction with an Attention Mechanism for Cyberattack Detection using Metaheuristic Optimization Algorithms (ODRAMCD-MOA). First, a min–max normalization model is used in the data preprocessing phase. Next, the Fruit Fly Optimization Algorithm (FOA) is employed for Feature Selection (FS). Subsequently, a hybrid of a Convolutional Neural Network (CNN) and a Bidirectional Long Short-Term Memory (BiLSTM) model with a Spatiotemporal Attention (STA) mechanism (CNN-BiLSTM-STA) is utilized for cybersecurity classification. Finally, the hyperparameters of the CNN-BiLSTM-STA method are optimized using the Augmented Red Panda Optimizer (ARPO). Comparative analysis shows that the ODRAMCD-MOA methodology achieves accuracy values of 99.79% and 99.03% on the ToN-IoT and BoT-IoT datasets, respectively, outperforming existing models.

Previous studies on automatic recognition of badminton strokes have shown persistent gaps, particularly uneven classification accuracy across stroke categories and suboptimal overall performance. To overcome these limitations, this study proposes a hybrid framework that combines handcrafted spatiotemporal features with a two-stage feature selection and weighted ensemble learning. From representative video frames, RGB-based descriptors (Histogram of Oriented Gradients (HOG), Histogram of Optical Flow (HOF), Motion Boundary Histogram (MBH)) and skeleton-based features (Range of Motion Index (ROMI) for spatial and Dynamic Time Warping (DTW) for temporal) are extracted. Dimensionality reduction is applied through Autoencoder compression followed by SelectKBest to preserve the most informative features. The refined hybrid features are then classified using a weighted soft voting ensemble that integrates Support Vector Machine (SVM), Logistic Regression (LR), Random Forest (RF), and Adaptive Boosting (AdaBoost) to enhance both accuracy and class-wise balance. Experimental evaluation on a badminton stroke dataset demonstrates the effectiveness of this approach, achieving 98.21% accuracy. The results highlight that the best-performing feature combination is ROMI, DTW, and HOF, confirming that hybrid handcrafted features with ensemble Machine Learning (ML) significantly improve robustness and stability, offering strong potential for practical applications in performance analysis, training systems, and sport analytics.

Gait analysis is essential for the diagnosis and monitoring of various biomechanical and neurological conditions. Clinical reliability is compromised due to imbalanced data and lack of transparency in existing models. This study uses vertical GRF features of patients to analyze gait on the publicly available GaitRec dataset. The proposed approach addresses the limitations of traditional models using Borderline SMOTE to balance the dataset and FedAvg-SVRG to stabilize gradient updates across dispersed nodes and enable federated learning with stable convergence without centralizing data. Enhanced LRPs are combined with LSTM to improve transparency in pattern recognition, and LIME offers instance-level interpretability. A deep learning model is used to improve the accuracy of gait classification, achieving 99.72% on single run execution and 99.45±0.14% for 10-fold cross-validation. The proposed deep learning framework holds significant potential for robust gait classification and clinical decision making.

Bridge inspections typically involve complex measurement criteria and take a long time. This study proposes a simpler inspection method that assesses the performance of reinforced concrete bridge girder beams. The method relies on crack length correlation with beam performance using theoretical approaches, Finite Element Model (FEM) simulations, and experimental testing. A reinforced concrete beam model is used to simulate similar flexural behavior to that of actual bridge girders. The results show that cracks form when concrete stress exceeds rupture stress, usually at 8%-12% of the ultimate capacity. Cracks grow until the steel reinforcement yields, which happens around 80%-90% of the ultimate capacity. Experimental crack length data at ultimate capacity align closely with FEM simulations, with a ratio of 0.92. In contrast, the results differ significantly from the theoretical predictions, with a ratio of 1.54. The study introduces six performance levels for reinforced concrete beams, linking crack length to beam performance. These parameters will help estimate bridge girder performance through visual crack assessments.