Predicting Electric Vehicle Adoption Intention Using Machine Learning: An AI-Driven Framework for Charging Infrastructure Planning in Karnataka, India

Forecasting electric-vehicle (EV) demand is important for charging-infrastructure planning, but aggregate sales models offer limited insight into which consumers are considering a purchase. The study's methodological contribution is a reproducible, leakage-controlled workflow that applies machine-learning classifiers to primary survey data to predict individual purchase consideration in a secondary-city context, as a complement to aggregate forecasts. A convenience and snowball sample of 175 respondents was drawn from Mysuru and surrounding districts of Karnataka, India; the analysis used the 162 non-EV owners (117 Yes/Maybe; 45 No). Logistic regression and random forest classifiers were compared using stratified five-fold cross-validation. Random forest achieved pooled out-of-fold accuracy of 0.784 (95% bootstrap CI: 0.722–0.846) and ROC-AUC of 0.843 (95% CI: 0.784–0.898), compared with 0.735 and 0.823 for logistic regression; repeated cross-validation gave slightly lower but consistent estimates. Attitude-related measures ranked highest, but their importance is predictive, not causal. Because the sample is small and non-probabilistic, the findings are a methodological proof of concept that requires external validation, not a tool ready for planning use.