A Hybrid Ensemble Learning Model for Electric Vehicles Performance Prediction and Classification

Abstract

Accurate forecasting of electric vehicles (EV) operations is a crucial part of improving battery performance and making sure that they are operating. This paper presents a hybrid ensemble learning algorithm that combines the three algorithms: Random Forest (RF), K-Nearest Neighbors (KNN), and Elastic Net (EN) to classify EV performance based on its operational parameters, which are current, load, life hour, life minute, distance and time. To evaluate it, the dataset was split into training (80% of the dataset) and testing (20% of the dataset) samples. Classification accuracy of individual models was 94.2 (RF), 88.6 (KNN) and 85.4 (Elastic Net). It was found that the proposed Hybrid Voting Classifier was better than standalone models with 97.1% accuracy. Besides, hybrid model had precision of 96.8, recall of 97.4, F1-score of 97.1, and AUC of $\mathbf{0. 9 8}$. The analysis of the confusion matrices showed that the misclassification rates are lower than the case with individual algorithms. The findings reveal that ensemble learning is very useful in increasing the Electric Vehicles performance prediction and assisting in intelligent battery management systems.