Traffic Flow Fuzzy Prediction Algorithm Based On Deep Learning

With the rapid development of the economy and the continuous improvement of automobile technology,traffic congestion is becoming more and more common,and relieving traffic pressure has become an urgent issue in the field of transportation.As a key link in the method of improving traffic capacity,traffic flow prediction is becoming a research hotspot in traffic science and intelligent technology.How to use traffic big data generated by information technology to conduct effective traffic flow prediction it is crucial.At present,deep learning has achieved many excellent results with its excellent learning ability.Aiming at the real-time and accuracy requirements of short-term traffic flow prediction,this paper combines ”traffic + big data + deep learning” to conduct a deep research,and proposes a novel traffic flow depth prediction method based on fuzzy theory.For the first time,fuzzy theory is used to introduce traffic accident uncertainty information into the convolutional neural network to assist prediction.The main work of this paper is as follow:1)This paper preprocesses the traffic flow data and extracts the uncertain information of traffic accidents hidden in the traffic flow data.For the problem of data missing in collected data,based on the periodicity of traffic flow,this paper fills the missing data with the traffic flow value of adjacent periods,and establishes an adjacent repetitive supplement method.In addition,in view of the characteristics that traffic accidents will affect traffic flow,the fuzzy theory is used to extract traffic accident information,which can effectively describe the traffic accident uncertain information implied in traffic flow data.2)Based on the pre-processed traffic data,this paper constructs a traffic flow depth prediction model based on the fuzzy theory.The model not only contains the spatial-temporal characteristics of traffic flow,periodically constructs the interior trend characteristic sequence of traffic flow as the input of depth network,but also uses the fuzzy reasoning to introduce traffic accidents uncertain information,constructs the accident feature sequence,and combines other weather information sequences as the input of depth network;Then,through the multi-layer convolutional layer and convolutional residual layer network learns the characteristics of input information to achieve effective traffic flow prediction.Furthermore,a novel traffic flow depth prediction method(F-CNN)is formed based on the prediction model,and the implementation process of corresponding prediction algorithm is introduced.Finally,the proposed prediction method is verified by the regional flow data and meteorological data of Beijing.3)This paper designs a traffic flow prediction prototype system.According to the formed traffic flow depth prediction method,a prototype system with data preprocessing function,traffic flow feature sequence construction function,deep learning model construction function,prediction model training and optimization function and future traffic flow prediction function is designed.And the system is applied to the intelligent transportation system,which can realize the construction scenario of future traffic conditions.