‏Spatial Data-Driven Traffic Flow Prediction Using Geographical Information System

Abstract

Today, traffic is one of the biggest problems of urban management. There are two general methods for traffic management, soft and hard methods. In the hard method, physical changes are applied to the road network, and in the soft method, the existing conditions are optimized. Traffic forecasting is one of the soft methods for traffic management. Traffic forecasting is usually done based on the time of existing traffic conditions, while the effect of location and neighborhood, which is one of the concepts of GIS science, is less seen in predictions. In this research, variables affecting traffic were first identified. Then, five machine learning methods were used to predict traffic on all city roads. KNN method was selected as the best one with accuracy and Kappa of 96.14% and 0.95 respectively. Finally, the prediction map was prepared by applying the superior model and Geographic Information System (GIS). One of the advantages of the traffic prediction map is easy for users and administrators to manage traffic.

Keywords

• geographic information system
• traffic
• prediction
• traffic map