Classification of potential electric vehicle purchasers: A machine learning approach

Among the many approaches towards fuel economy, adoption of electric vehicles (EV) may have greatest impact. However, existing studies on EV predict very different market evolutions, which causes a lack solid ground for strategic decision making. New methodological tools, based Artificial Intelligence, might offer perspective. This paper proposes supervised Machine Learning (ML) techniques to identify key elements in adoption, comparing ML methods classification potential purchasers. Namely, Support Vector Machines, Neural Networks, Deep Gradient Boosting Models, Distributed Random Forests, and Extremely Randomized Forests are modeled utilizing data gathered users’ inclinations EV. Although with polynomial kernel slightly outperforms other algorithms, all them exhibit comparable predictability, implying robust findings. Further analysis provides evidence that having only partial information (e.g. socioeconomic variables) has significant negative impact model performance, synergy across several types variables leads higher accuracy. Finally, examination misclassified observations reveals two well-differentiated groups, unveiling importance profiling purchaser marketing campaigns as well public agencies seek promote adoption.