Predicting the impact of text-reading using decision trees

Abstract

Various road safety analyses prove that cell phone usage cause driver distraction which, in turn, has become a leading cause for crashes. Various studies have focused on different cell phone operations such as hand-held or hand-free conversation, number dialing and text writing and reading and examined how they affect driving performance. Research efforts have been also placed on investigating the effects of sociodemographic characteristics on distraction and related them to the reaction of the drivers under distraction and the resulting speed, lane changes, lateral placement, deceleration, incidents and many other variables. The primary aim of this paper is to implement a decision trees approach in predicting the degree of influence of text reading on driving performance and associate it with self-reported behavioral and sociodemographic attributes. Data were based on a sample of 203 taxi drivers in Honolulu, who drove on a realistic driving simulator. Driving performance measures were collected under non-distraction and text-reading conditions. Among them, line encroachment incident and maximum driving blind time changes were used in combination with sociodemographic characteristics (gender, age, experience, educational level, race) and behavioral constructs (past behavior, behavior, behavioral beliefs, control beliefs, risk appreciation and descriptive norms) and decision trees were built. The analysis revealed that important predictors for maximum driving blind time changes are sociodemographic and past behavior attributes. The accuracy of the prediction increases in the case of line encroachment incident changes, with the addition of behavioral beliefs, control beliefs, risk appreciation, descriptive norms and past behavior. © Springer Nature Switzerland AG 2019.