Reducing Student-Instructor Mismatches in Driving Schools Through Compatibility Analysis

Abstract

Matching students with suitable instructors is critical for learning effectiveness in driving schools, yet most schools rely on manual, experience-based allocation. This paper presents a real-world case study at a Japanese driving school analyzing student-instructor compatibility using operational data. We integrate 5-point and 2-point evaluation data, perform exploratory analysis connecting aptitude-test personality patterns to instructor ratings, and build a LightGBM classification model to predict mismatches. Rather than recommending optimal pairs, we focus on reducing mismatches by avoiding low-compatibility pairings while respecting equal distribution, enabling instructors to maximize their strengths and improving student outcomes.