Analysis of the traffic flow and accidents with big data algorithms

In terms of social and individual living conditions, traffic has been one of the most significant problems of today's world. As a result of increasing population and vehicle number, existing traffic problems increase exponentially. Because of this reason, time wasted in traffic and possibility of having an accident increases with the rate. This thesis intends to relieve the vehicle traffic and reduce the probability of occurrence of risks to minimum or prevent in advance. In this context, big data infrastructures and algorithms are used in order to predict accident, traffic congestion, and road works, without notice works, which are named as "event". To predict all these events, this work takes advantages of artificial neural networks and models optimized for traffic congestion problems. The main purpose of the study is to predict real-time traffic events or congestion probability on time and offer related precautions to be taken. In the context of the thesis, the data from the traffic sensors has been collected for one year and used to form a model from the pilot area of Istanbul, which is known as one of the most crowded traffic on the world. As a result of the study and the model, traffic events has been detected with an accuracy rate of 97%. The events have been detected taking in the consideration of the data velocity. Based on the detections, the thesis has aimed transferring the traffic to another road, notifying the emergency units and proving the traffic security and resolving the traffic congestion with the help of the change in signalization. The study has been constructed with the 2015 traffic data from Istanbul Metropolitan Municipality Traffic Unit and the model formed by this data has been considered to be developed and accommodated with the changing conditions.