Symbiosis University of Applied Sciences

This study presents a structured framework for gear fault classification under varying speed and load conditions using statistical signal features, classical machine learning (ML), and quantum-inspired feature embedding. Time- and frequency-domain features extracted from vibration and torque measurements are employed to classify healthy and faulty gear states, including micropitting, pitting, wear, and tooth flank fracture. Classical ML models provide strong baseline performance, while quantum-inspired feature embedding improves class separability for overlapping and noise-sensitive fault conditions. The results indicate that quantum feature mapping can enhance interpretability and feature-space separability in complex diagnostic scenarios, although classical models remain competitive for well-separated fault classes.

Fraud detection in financial analytics is a critical challenge that demands advanced deep learning techniques for accurate and efficient identification of fraudulent activities. This paper explores the application of Long Short-Term Memory (LSTM) networks, a powerful Recurrent Neural Network (RNN) architecture, to detect and prevent financial fraud. LSTM's ability to capture sequential dependencies in transactional data enables the model to learn complex patterns and anomalies effectively. Additionally, we leverage AutoML to optimize hyperparameters and model selection, enhancing detection accuracy while minimizing manual intervention. The proposed system is evaluated on real-world financial datasets, demonstrating superior performance in identifying fraudulent transactions with high precision and recall. Our results suggest that integrating LSTM with AutoML significantly improves fraud detection in dynamic financial environments.

This work offers a fresh hybrid method for heart disease prediction using unsupervised, semi-supervised, and supervised learning methods to improve accuracy, especially in situations with minimal labelled data. To efficiently use both labelled and unlabelled data, the method integrates data preparation, feature engineering, K-means clustering, label propagation, and ensemble learning. The semi-supervised learning phase uses label propagation to infer labels for unlabelled data points; the unsupervised learning phase uses K-means clustering to identify patient groupings. Modern ensemble approaches include Random Forest and stacking techniques, which help capture complicated interactions between features, and are included in the component of supervised learning. Extensive studies employing real-world clinical datasets have shown that the suggested hybrid approach regularly outperforms the conventional techniques. The paper also covers ethical issues in the creation and implementation of AI-based medical prediction models and investigates model interpretability. Better clinical decision-making resulting from the increased accuracy, resilience, and interpretability of this method should improve patient outcomes and more effective allocation of healthcare resources. Moreover, the flexibility of the framework to different medical diagnostic activities creates fresh research opportunities in several spheres of healthcare, thus supporting the disciplines of medical artificial intelligence and machine learning.

In this study, the hand gestures prediction for electromyography (EMG) signals classification is provided. EMG signals which are electrical stimulations of muscles were measured, processed and analyzed to achieve discriminative features in time, frequency and time-frequency domain. To improve a better result of the classification, a number of classifiers namely, Support Vector Machine (SVM), k-Nearest Neighbors (k-NN), Random Forest (RF) and Artificial Neural Network (ANN) have been compared with each other with the objective of hybrid feature combination to enhance the accuracy of the result of the classification. As it turned out, ANN proved to be the most precise in classification (96.2 percent), followed by SVM (94.6 percent), RF (92.7 percent) and k-NN (89.4 percent). These findings suggest that hybrid feature extraction works better in terms of discriminative power and nonlinear learning capacity of ANN can be employed to extract the complicated and non-stationary nature of EMG signals. Overall, this research paper indicates the potential of hybrid feature extraction in conjunction with sophisticated learning architectures to enhance performance, robustness and flexibility of EMG-based hand gesture recognition systems to be applied in the field of prosthetic control, rehabilitation robotics as well as human-machine interfaces.

In this study, we introduce HyperGraphViz, a Hypergraph Neural Network (HGNN)-based AI framework designed for context-aware email communication analysis using the Enron Email Dataset. Traditional graph-based models fail to capture multi-recipient interactions, limiting their ability to detect fraud and insider threats. HyperGraphViz addresses this limitation by leveraging hypergraph structures to model complex, hierarchical relationships within corporate networks. The model employs Hypergraph Convolutional Networks (HGCN) for feature extraction and anomaly detection, coupled with natural language processing (NLP) techniques for analyzing email content. The experimental results demonstrate that HyperGraphViz outperforms conventional models, achieving an accuracy of 91.3