Strategic Forecasting for Electric Vehicle Sales: A Cutting Edge Holistic Model Leveraging Key Factors and Machine Learning Technique

This study aims to explore the influential factors affecting electric vehicle (EV) sales through an extensive literature review and identify these factors' importance in EV sales forecasting. The research also seeks to validate an effective forecasting model for future EV sales. Twelve significant factors influencing EV sales were identified from the literature. Further, expert opinions and the DEMATEL method were used to prioritize these factors, and their influential weights were incorporated into the dataset for EV sales forecasting. Additionally, multiple machine learning techniques, including ARIMA, XGBoost, Random Forest, and LSTM, were employed to predict future EV sales. The study highlights twelve significant factors affecting EV sales, emphasizing historical EV sales, the share of EVs sold, and electricity demand as key causal factors. DEMATEL results indicate that among the nine causal factors, historical EV sales and the share of EVs sold are the most impactful. Furthermore, the Random Forest technique was found to perform the best among the machine learning methods, effectively capturing complex relationships and enabling accurate predictions. The proposed machine learning-based model, incorporating influential weights, offers a robust framework for forecasting EV sales using historical sales data. This model aids policymakers and industry stakeholders in planning EV infrastructure and electricity demand. It supports the EV supply chain and promotes a greener future by facilitating the planning and popularization of EVs and widespread electric mobility.