Fitness-To-Drive Assessment of Older Drivers Based on Multi-Classification Predictive Models

Abstract

Aging has become a global issue, which is accompanied with the increase of older drivers on the road. To reduce the road safety problems caused by older drivers, it is necessary to assess their fitness-to-drive. In this study, by identifying three categories of older drivers from on-road driving test (i.e., "fit to drive", "undetermined" and "unfit to drive"), six indicators from their functional ability tests and the simulated driving test were extracted, and three machine learning methods, i.e., decision tree, random forest and support vector machine, were applied to conduct the fitness-to-drive assessment among a number of older drivers. The result showed that the support vector machine achieved the best performance, with the highest accuracy rate over 80%, which implies the feasibility of using the proposed assessment procedure as an alternative way to the on-road test for the fitness-to-drive assessment for older drivers.