Traffic Prediction and Planning Based on Time Series Forecasting and Simulated Annealing Algorithm

This paper presents a comprehensive approach for traffic prediction and planning, leveraging time series analysis and Simulated Annealing (SA) algorithm. Initially, traffic flow data undergoes preprocessing to identify trends in different directions, distinguishing peak and off-peak periods. Using K-means clustering, similar traffic flow periods are categorized into three clusters for temporal segmentation. Bayesian inference is then employed to estimate straight and turning probabilities across four traffic directions. Subsequently, we develop an optimized signal control model and an intersection capacity model using planning methodologies, establishing phase-specific traffic flow models, calculating vehicle volumes, and formulating average speed distribution models at intersections with defined constraints, solved using Lingo. Finally, a queueing theory-based model is constructed to preprocess and organize data, determining cruising vehicles. Assuming a Poisson distribution for vehicle arrivals, parameters are estimated, and the Simulated Annealing algorithm is utilized to determine an optimal parking requirement of 42 spaces.