Content-based Recommendation for Traffic Signal Control

Traffic signal control is an effective way of solving urban traffic problems by providing appropriate signal control plans for various intersections. Essentially, the aim of Traffic Signal Control is to find the best matching timing plans to current traffic conditions. Inspired by recommendation technology, we regard traffic conditions as users, timing plans as items, and traffic indicators like delay time are regarded as the ratings that users give to items. By means of Content-based Recommendation technology and k-Nearest Neighbor method in Recommendation Systems, we first find the similar traffic conditions according to the characteristics of traffic conditions. Then the matching degree between current traffic conditions and various timing plans can be predicted by analyzing the history data of selected similar traffic conditions. What's more, Artificial Transportation Systems method was applied to recommend and sort the timing plans for various traffic conditions in this paper. With normalized Discounted Cumulative Gain, which is a measure of ranking quality, was chosen as the performance indicator, we conducted the experiments in Paramics. The results showed that the strategies based on our method outperform the classic Webster method.