Traffic Flow Prediction On Factorization Machine

In recent years,the infrastructure construction of China's urban traffic network has developed rapidly.However,with the development of the economy,the road construction measures are far from being able to meet the growing travel needs.As a result,the traffic network of many large cities often suffers from traffic jams and even faces widespread disaster.Traffic flow forecasting is an important research area of Intelligent Transportation Systems.Accurate traffic flow forecasting can achieve applications such as traffic signal control,route guidance,and accident detection in intelligent transportation systems.It is a very important basic theory in the field of intelligent transportation.Through the analysis and prediction of traffic flow,it can effectively help the city to conduct intelligent traffic guidance work and provide convenience for people's travel.As a machine learning method based on matrix decomposition,factorization machine can solve practical problems such as non-linear,high-dimension,and high sparsity.It is a cutting-edge technology in the field of complex nonlinear science and artificial intelligence,and is very suitable for solving complex cities.The problem of forecasting traffic flow of road network realizes the intelligent control of urban road traffic and effectively relieves the pressure of urban road traffic.Based on the analysis of traffic flow forecasting based on factorization machine,this paper based on the analysis of urban road network structure and traffic flow data,carried out in-depth research and discussion on urban road network traffic flow prediction methods.And based on the actual traffic flow data construction method model,with practical application of feasibility.The main work is as follows:(1)Through the analysis of traffic flow data and the identification and processing of errors and missing data,reduce the impact of noise on the prediction process,and lay the foundation for the establishment of the next-generation traffic flow prediction model;(2)According to the basic principle of the factorization machine,the traffic flow prediction model based on the factorization machine is studied and adopted,and the prediction evaluation index is used to determine the accuracy of the prediction result.Experiments show that the factorization machine model is a feasible and effective traffic flow prediction model.(3)Optimize the input samples of the factorization machine.The quality of the input samples directly affects the accuracy of the prediction results.In this paper,we use the auto-encoder algorithm to optimize the input samples.Through training,the initial value of the input sample is optimized,and then the model is used for prediction.Experimental results show that using an auto-encoder to optimize the input samples can effectively improve the prediction accuracy.