Gradient Boosting Approach for Traffic Flow Prediction Using CatBoost

Traffic congestion has been a worldwide issue that has a massive impact on our daily lives. The increasing number of vehicles intensifies traffic congestion, especially in urban areas. Congestion identification is the first step in choosing the right way to avoid this situation. Predicting the flow of traffic has gotten much attention recently as it is an important step in preventing and reducing traffic congestion. However, it seems that there are some important factors such as no. of intersections on the street, no. of commercial places near the streets, and structure of street which contribute to the traffic congestion. In this paper, we propose an approach for predicting the traffic congestion level of 37 streets using the structural information and the previous data of traffic congestion on those streets. Our proposed system classifies congestion into 3 levels (0,1,2) where congestion increases with levels with an accuracy of 97%.