A Study On Short-term Traffic Flow Forecasting Method Based On Big Data Mining Technology

Accurate and real-time traffic flow forecasting is an important basis for traffic scheduling,traffic light timing and traffic diversion and induction.Short-term traffic flow forecasting is usually carried out by using the similarity and periodicity of traffic flow,but due to other accidental factors such as the environment,the traffic flow is characterized by irregularity and complexity.Short-time traffic flow forecasting needs to explore the hidden rules in this irregularities as far as possible,so as to obtain more accurate forecasting results,the paper mainly does the following work.(1)The paper proposed a new method for solving spatio-temporal correlation coefficients.On the basis of the study of traffic flow characteristics and spatio-temporal correlation of traffic flow,it was found that the traditional spatial correlation coefficients was calculated by Pearson function,which lacked the traffic characteristics for traffic system.Aimed at this problem,the paper redefined the distance weight matrix,and proposed the lane weight matrix based on lane width.The new solution method of correlation coefficient was redefined by using the new distance weight matrix and lane width matrix.(2)The paper modified the standard grey wolf optimizer algorithm.The standard grey wolf optimizer algorithm could not distinguish the global search and local search very well,and was easy to fall into local extremum.for this reason,the convergence factor in the grey wolf optimizer algorithm is improved in this paper.The Sigmoid function is introduced to make the convergence factor adaptively decreasing;and in the position renewal formula of the individual of the wolf group,the inertial weight is added,and the size of the inertial weight is adjusted so that the grey wolf optimizer algorithm has the ability to jump out of the local extremum.(3)Based on the spatio-temporal correlation characteristics of traffic flow,the improved grey wolf optimizer algorithm was used to optimize the BP neural network model to forecast the short-time traffic flow forecasting.The paper built a large data cluster of MATLAB,and Guiyang's traffic bayonet data was used to constructing the state space vector constructed,and an example is verified,and the prediction accuracy is analyzed by mean absolute error,average absolute percentile error and mean square root error.The research of this paper shows that the improved grey wolf optimizer algorithm based on space-time is used to optimize the BP neural network model,the prediction results are more accurate and the processing speed is faster,which shows the short-term traffic flow prediction based on big data mining technology is reliable,effective and has a good application prospect.