Naive Bayes Classifier Based Driving Habit Prediction Scheme for VANET Stable Clustering

Vehicular ad hoc network (VANET) is expected one of promising network forms for intelligent transportation system which supports road safety applications, in-vehicle entertainment and arriving automatic driving. Establishing and maintaining stable connections in VANETs are challenging on account of the high mobility of vehicles, dynamic vehicle topology, and time-varying vehicle density. Clustering can provide scalability and reliability for VANETs by grouping vehicles with hierarchical structures. However, keeping cluster stable became a hard nut to crack due to high vehicle speed and unpredictable driving pattern. Recent rapid development of artificial intelligence (AI) provided an innovative solution for this situation. In this paper, a Naive Bayes Classifier based driving habit prediction scheme for stable clustering is proposed, briefly named NBP. According to driving speed and overtaking decisions, vehicles are classified into two alignments with different driving habit. Specifically, Naive Bayes classifier perform driving habit prediction through several relative independent factors, such as relative velocity, vehicle type, number of lanes traveled. The cluster head candidates will be chosen from alignment with mild driving pattern which will benefit for stable clusters. Combined with clustering design, the proposed method has been proven effective for stable clustering in VANET based on the real data of highways in California.