Artificial intelligence-based delay prediction models for signalized intersections in urban areas

Intersections are significant road elements for traffic safety and road capacity. Furthermore, intersections have serious impacts on travel time. Time lost due to deceleration and stopping manoeuvres increases travel time and causes delays. Various factors affect the intersection delay. However, the effects of public transportation on delays also need to be investigated. This study focuses on these impacts on delays. The delays at four-legged-signalized-intersections were studied in the city-center-of Denizli, Türkiye. Intelligent Transportation System (ITS) was used to obtain information from both the traffic and public transportation systems, and a common database was built by cleaning and processing data. Multiple linear regression and artificial intelligence techniques were used to predict delays and then these methods were compared. The findings show that the k-nearest neighbor and artificial neural network give the best results with symmetric mean absolute percentage error values of 14.3% and 15.31%, respectively. In addition, the root mean square errors of these methods were found to be 10.47 and 10.42 s, respectively.