CAFWE: A Feature-Weighted Ensemble Classifier for Predicting Performance of Copper-Aluminum Composite Conductors

Abstract

Copper-aluminum composite conductors are suitable for modern electrical applications due to their high electrical conductivity and lightweight nature; however, predicting their performance in changing environmental and electrical conditions is difficult and expensive using conventional techniques. Problem Statement: Previous models ignore the combined effects of copper-aluminum ratio, tensile strength, temperature, and current load; furthermore, metrics such as accuracy and F1-score cannot accurately reflect physical performance characteristics. A new Copper-Aluminum Feature-Weighted Ensemble (CAFWE) classifier was developed using the CopperAluminum_WirePerformance_Dataset containing 5000 samples with 10 input features and one target output (performance_level). The model integrates three base learners — Linear Classifier, k-Nearest Neighbor (k = 5), and a Decision-Stump classifier — combined through a weighted voting mechanism that assigns higher weights to copper_percentage, tensile_strength, and electrical_conductivity based on feature-sensitivity analysis. The dataset was partitioned using an 80:20 stratified split, and all results were averaged over five repeated experiments to ensure stability. Five domain-specific evaluation metrics were introduced: Conductivity Accuracy (CA), Strength Reliability (SR), Temperature Adaptation Score (TAS), Load Prediction Stability (LPS), and Composite Material Alignment (CMA), enabling alignment with real-world engineering behavior. Results: Across five independent training runs, CAFWE achieved consistent performance, with mean scores of CA = 93.5%, SR = 91.2%, TAS = 89.8%, LPS = 90.6%, and CMA = 92.4%, demonstrating superior predictive reliability under varying material and operational conditions. Feature importance analysis confirmed copper_percentage and electrical_conductivity as the most influential contributors to final predictions. The CAFWE model accurately and interpretably predicts copper-aluminum conductor performance; and provides a scalable framework to optimize hybrid material design for smart grid applications.