AI-Driven Strategies for Predicting the Adoption and Impact of Clean Energy Technologies in the US Automotive Sector

Abstract

The rapid shift towards clean energy technologies requires strong, data-driven tools to forecast adoption trends and evaluate their various impacts. This research presents a comprehensive AI framework that integrates diverse datasets, including vehicle registrations, emissions records, research and development investment figures, demographic and socioeconomic indicators, and logs of policy legislation, into a unified analytical platform. Through systematic feature engineering, we create an Adoption Index that combines factors such as income, fuel prices, and registration growth, alongside a Policy Incentive Score, metrics for infrastructure density, and temporal markers (quarter, year) to identify hidden drivers of clean technology adoption. Exploratory data analysis reveals regional and temporal patterns of adoption, emphasizing the connections with socioeconomic status and legislative factors. We employ a range of predictive and analytical models: ensemble regressors (Random Forest, XGBoost), LSTM time-series networks for forecasting trends, K-Means and DBSCAN for regional segmentation, and causal-inference methods (DoWhy) to assess policy effectiveness. The models are evaluated using R², RMSE, and MAE for regression tasks; Silhouette Score and Davies-Bouldin Index for clustering; and Average Treatment Effect (ATE) for policy impact analysis. Our XGBoost regressor achieves an R² of 0.89 and an RMSE of 5.7% in predicting adoption rates, while the LSTM models capture temporal dynamics with a 6.1% RMSE. Clustering reveals three distinct adoption archetypes, and causal analysis indicates that doubling tax credits could increase adoption by 20% (ATE = 0.20). These findings demonstrate the effectiveness of integrated AI strategies for forecasting and evaluating policies in the transition to automotive clean energy.