Driving Behavior Evaluation Model Base on Big Data From Internet of Vehicles

Abstract

This article attempts to evaluate whether a driving behavior is fuel-efficient. To solve this problem, a driving behavior evaluation model was proposed in this article. First, the operating data and fuel consumption data of five trucks were obtained from the vehicle networking system. Four characteristic parameters, which are closely related to fuel consumption, were extracted from 19 sets of vehicle operating data. Then, K-means clustering combined with DBSCAN was adopted to cluster the four characteristic parameters into different driving behaviors. Three types of driving behavior were labeled respectively as low, medium and high fuel consumption driving behavior after clustering analysis. The clustering accuracy rate reached 79.7%. Finally, a fuel consumption-oriented driving behavior evaluation model was established. The model was trained with the labeled samples. The trained model can evaluate the driving behavior online and gives an evaluation of whether the driving behavior is fuel-efficient. The test results show that the prediction accuracy rate of the proposed model can reach to 77.13%.