Analysis of Istanbul traffic with data mining and machine learning methods: D100 highway application

For the prediction of traffic measures in the near future based on the traffic problem, current and future traffic data, Improving the traffic planning and management of users and decision makers in Intelligent Transportation Systems has recently become a very important issue to cope with the traffic problem. Thanks to the Intelligent Transportation Systems, it has enabled the formation of data-oriented models as a result of the formation of a large amount of traffic data. There is an increasing interest in the prediction of traffic measures by modeling big data-driven complex scenarios with data mining and machine learning methods. In this study, in the proposed traffic analysis modeling, the parameters affecting the traffic of the region to be investigated were determined by statistical methods, and their relations and effects with each other were revealed. The relationships between traffic density, and parameters of hour, day, month, season, year, weather events, accident situations were determined. Obtained parameters were modeled with machine learning methods and traffic density analysis and traffic incident analyzes were performed. In this model, the aim has been an important study for the realization of a spatially specific traffic warning system, since each region has its own parameters that affect the traffic problem. The traffic density on the Anatolian side, the European side and the 15 July Martyrs Bridge lines of the Istanbul D100 highway and the parameters affecting the traffic incidents were determined by statistical methods and analyzed. Models of traffic density and accident events were created with Bayesian network and Artificial neural network methods. Thanks to these models, the infrastructure of the early warning system has been created for traffic density situations and traffic incidents that may occur specific to the region.