Development of Data Mining Methodologies to Advance Knowledge of Driver Behaviors in Naturalistic Driving

This article presents data mining methodologies designed to support data-driven, long-term, and large-scale research in the areas of in-vehicle monitoring, learning, and assessment of older adults' driving behavior and physiological signatures under a set of well-defined driving scenarios. The major components presented in the article include the instrumentation of an easily transportable vehicle data acquisition system (VDAS) designed to collect multimodal sensor data during naturalistic driving, an ontology that enables the study of driver behaviors at different levels of integration of semantic heterogeneity into the driving context, and a driving trip segmentation algorithm for automatically partitioning a recorded real-world driving trip into segments representing different types of roadways and traffic conditions. A case study of older driver arousal levels in various driving contexts using the proposed methodology is presented to demonstrate that the proposed data mining infrastructure and methodologies are effective in analyzing driver behaviors through recorded real-world driving trips.