Traffic Flow Forecasting Model with Density Based Clustering Algorithm

Population, expansion of the residential areas and increasing number and density of vehicles affect city traffic and bring front the transportation as a problem. Advanced traffic management systems are always required to control, regulate and manage city traffic. Many researchers have done a number of studies to regulate, develop and improve the systems, and the most important of them are about the forecasting of traffic flow, route, driving time, etc. In this study, without spatial road information, traffic flow forecasting has been done with the forecasting approach based on density based OPTICS algorithm and considering the position information and angular directions data that obtained from the vehicles. When developing this model, firstly, all, vehicle data of our country were filtered to determine latitude-longitude coordinates including Ankara-Istanbul highway, location and angular information were scaled, the data was clustered then by OPTICS clustering algorithm with different adjacent values, and it was seen that vehicles can be clustered at the determined proximity/distance with the developed model. The obtained results indicating that, without road information, vehicles can be grouped at intended scale, road condition coloring and traffic flow forecasting can be made with the proposed model. It can be concluded that this study might guide new studies in traffic flow forecasting in future studies.