A Novel Data-Driven Multi-Branch LSTM Architecture with Attention Mechanisms for Forecasting Electric Vehicle Adoption

Electric vehicle (EV) adoption has surged in recent years, underscoring the need for accurate forecasting models to inform policy and industry planning. Traditional forecasting approaches often struggle to capture the complex, nonlinear patterns in EV adoption data, motivating the use of deep learning methods. This paper proposes a novel multi-branch Long Short-Term Memory (LSTM) architecture enhanced with attention mechanisms to forecast EV adoption trends. We review the state-of-the-art in EV adoption forecasting, including classical time-series models and modern deep learning techniques. We then analyze current multi-branch LSTM and attention-based models, highlighting their advantages in capturing multifaceted data influences. A critical comparison of alternative methods is presented along with their performance metrics and limitations. Building on identified gaps—such as scalability and integration of heterogeneous data—we outline the scope and methodology for the proposed model. The multi-branch LSTM with attention is expected to improve forecasting accuracy and interpretability by simultaneously learning from multiple data streams (e.g., historical adoption, economic indicators) and focusing on the most relevant temporal patterns. Figures and tables are included to illustrate EV adoption trends and comparative performance of forecasting models.