Deep Learning-Based EV Adoption Forecasting

Sri Lanka's EV market is expanding faster than the tools used to understand it. Traditional econometric forecasting fails to capture the non-linear interplay of price sensitivity, charging-infrastructure gaps, import policy uncertainty, and the distinct value-driven purchasing behavior of Generation Z — the demographic that will define the market's next decade.

This study built a hybrid deep learning framework to close that gap. A primary survey of 643 Sri Lankan Gen Z respondents feeds a Deep Neural Network combined with discrete choice modeling and LSTM-based temporal forecasting, producing consumer segmentation, purchase-probability prediction, and market-growth trajectories in a single pipeline. Model performance was validated using F1-Score and ROC-AUC across held-out test sets.

The output gives automotive manufacturers, EV-ecosystem investors, and policymakers a data-grounded view of who adopts, when, and why — replacing intuition-led market entry decisions with quantified demand signals for an emerging market where global adoption curves don't transfer cleanly.