Optimizing Multi-Class Fault Detection in Digital Circuits Using a Hybrid Ensemble of Random Forest and XGBoost

Fault detection in digital circuits, especially for multiclass classification problems, is challenging due to issues like class imbalance and the demand for high accuracy. The aim of this study is to design a hybrid ensemble approach where Random Forest and XGBoost are combined to enhance classifying ability, reduce overfitting, and improve resilience to noise. While Random Forest offers stability by aggregating multiple decision trees, XGBoost provides targeted error correction with its iterative boosting approach, helping to improve overall fault detection. Even though there are some computational trade-offs, the hybrid model is effective in combining these strengths, achieving better accuracy and robustness compared to standalone models. Further, the hybrid model illustrated beautifully in handling of imbalanced data, making it appropriate and suitable for real-time and scalable fault detection tasks. Models were evaluated using "F1-Score, Precision and Recall", demonstrating a robust performance. The Proposed Hybrid model achieved an accuracy of 91%, surpassing Random Forest at 88% and Decision trees at 85%. Further research may focus on optimizing its performance for resource-constrained settings or more complex fault detection conditions.