Data analytics and optimization in traffic applications

Transportation is an indispensable need of all people in society. It is becoming more and more difficult to travel by foot or by car in or out of the city due to the increase in population and number of vehicles. The material and moral costs caused by the time lost due to transportation become difficult to bear. Traffic stress due to traffic congestion, violations and accidents also create a negative ground for many people's business and domestic life; even traffic stress is thought to reduce employee productivity. Therefore, the size of the cost due to the loss of time in traffic is quite high. To prevent and reduce the severity of these damages, optimization, simulation, and prediction studies are carried out for finding optimal routes in traffic, identifying traffic bottlenecks, analyzing accidents, etc. In this study, data analytical studies are conducted for the problems related to traffic. Firstly, a novel and hybrid model based on Artificial Neural Network is proposed for forecasting the traffic flow which is considered as a time series. Compared with the well-known time series forecasting methods, the proposed hybrid algorithm is found to be more successful. In addition, the accidents in Eskisehir are analyzed, probability distributions of these accidents are determined and data mining methods are used for accident prediction. Lastly, a dynamic programming approach based on the Markov Decision Process is used to determine the shortest route offline in uncertain and dynamic traffic networks by making use of forecasts and obtained probability distributions. When the applied holistic approach is tested on the generated problems, better results are obtained than the deterministic approach.