Short-term Traffic Flow Forecasting Based On FCM-EFCNN Combined Model

With the continuous development of society and economy,people's demand for transportation is also changing gradually.Accurate and real-time short-term traffic flow prediction can not only provide data support for the rationality analysis of road network structure,but also make people travel more convenient and efficient.Therefore,the prediction of short-term traffic flow is of great significance to the improvement and development of the traffic system.Short-term traffic flow forecasting can be regarded as a multi-factor space-time series forecasting problem.However,the existing forecasting methods in this field are still limited by the following data characteristics and model structure: 1)traffic flow data has complex characteristics and properties,and a single forecasting model often needs strong feature extraction ability and complex calculation process,and the construction process is very complex.Complexity;(2)insufficient pretreatment of traffic flow data often results in inadequate use of effective information;(3)model is insensitive to abnormal traffic flow data caused by emergencies,resulting in lower accuracy of prediction results.In view of the above problems,the main research contents of this paper are as follows:(1)To overcome the shortcomings of traditional forecasting models,such as inadequate use of spatial and temporal information of traffic flow and insensitivity to abnormal data changes caused by emergencies,an error feedback convolution neural network forecasting model is proposed and applied to short-term traffic flow forecasting.The improved model utilizes the advantages of convolutional neural network such as local perception and weight sharing to efficiently analyze traffic flow spatio-temporal data.At the same time,the error feedback mechanism of the network is optimized,so that the model can better identify the characteristics of abnormal traffic flow data and improve the accuracy and adaptability of the prediction model.(2)To overcome the shortcomings of single model in short-term traffic flow prediction,a combined FCM-EFCNN model is proposed and applied to short-term traffic flow prediction.The combined model uses the fuzzy clustering algorithm to divide the traffic flow pattern,and constructs the combined model according to the result of the division.The error feedback convolution neural network is used as a sub-model to predict the short-term traffic flow.Relevant experiments show that the FCM-EFCNN short-term traffic flow combination forecasting model proposed in this paper is more in line with the actual situation in the division of traffic flow patterns.At the same time,the combination forecasting method can reduce the difference of data received by the sub-model to a certain extent,so as to better identify the characteristics of traffic flow data and further improve the prediction accuracy of the model.