International Journal of Intelligent Data and Machine Learning

Abstract

Predictive modeling in the automotive industry often involves analyzing customer behavior to anticipate events such as vehicle purchases, service visits, or campaign responses. However, when working with imbalanced data—such as rare events like luxury vehicle purchases or high-ticket service upgrades—over-sampling techniques are commonly used. These techniques introduce bias into the sample, requiring adjustments to predicted probabilities to reflect the true population proportions. This paper explores the methodology of adjusting predicted probabilities using prior probabilities and demonstrates its application in automotive propensity models.