A DECISION TREE BAGGING APPROACH FOR EV BATTERY DEGRADATION PREDICTION

The adoption of Electric Vehicles (EVs) is accelerating globally, with sales exceeding 14 million units in 2023 a 35% rise from the previous year. Research highlights that battery performance and energy forecasting contribute to nearly 45% of operational inefficiencies in EVs. Moreover, optimized charging strategies can reduce energy wastage by up to 30%. However, most current systems depend on manual estimation or simplistic rule-based methods for battery monitoring, which often yield inconsistent results due to human error, limited adaptability to real-time data, and an inability to model the non-linear degradation of batteries. To address these challenges, this study introduces a machine learning-based predictive framework that utilizes Bagging with Decision Tree Regressor (DTR) to estimate the Remaining Useful Life (RUL) of EV batteries. The approach begins by collecting high-dimensional EV battery datasets and applying thorough preprocessing steps such as feature scaling, outlier elimination, and unique value assessment. Baseline models like Support Vector Regressor (SVR) and Deep Neural Network (DNN) are implemented for performance benchmarking. The proposed ensemble technique applies Bagging, training multiple DTRs on diverse bootstrap samples and averaging their outputs to enhance accuracy and reduce variance. This predictive system is embedded within a Django-powered web application that supports user interaction, realtime predictions, and on-demand model training using uploaded battery data. The resulting platform is scalable, interpretable, and delivers highly accurate forecasts, enabling better EV charging strategies and efficient battery lifecycle management.