Segment Based Highway Traffic Flow Prediction in VANET Using Big Data Analysis

Globally, abnormal highway traffic flow is one of the significant issues that directly or indirectly impact humans' well-being. It affects the economy and the environment significantly. Vehicular Ad-hoc Network (VANET) and Intelligent Transportation System (ITS) can improve the traffic flow of vehicles on highways by utilizing Vehicle to Vehicle (V2V) communication technology that enables vehicles to share and update traffic information in real-time. This paper proposes a system for predicting traffic flow in a VANET environment where vehicles on a highway segment form a cluster with a lead vehicle serving as the cluster head. The cluster head uses beacon information received from the vehicles in its cluster to determine the occupancy and the average speed in the corresponding segment and send this information to the Roadside Unit (RSU) to be used in the prediction of abnormal traffic flow. In the absence of real-life VANET datasets, we have successfully created one for this system by utilizing real-life traffic data collected by the Florida Department of Transportation District 4 (FDOT-D4). The created VANET dataset is used in building, training, and validation of several machine learning models, and the models' performance is evaluated in terms of accuracy and area under the receiver operating characteristic curve (AUC-ROC). The top two performing models are presented in this study.