Research On The Passenger Flow Forecast Of Urban Rail Transit Based On Improved Least Square SVM

With the continuous expansion of China’s urban scale,the increasing demand of urban transport has become an important issue that restricts social development.Urban rail transit has become the most important part of urban public transport system and is responsible for the main transport tasks of urban passenger flow.As it is of great significance to predict urban rail transit passenger flow,urban rail traffic passenger flow is the basis of the preliminary planning of the rail transit and operation department traffic scheduling in the later period.Based on the measured passenger flow data of the Sichuan Gymnasium Station of Metro Line 1 in Chengdu,in-depth analysis of the inbound passenger flow,early-peak passenger flow and sudden passenger flow forecasting problems is achieved in the dissertation.The main research contents are as follows:Firstly,in the dissertation,the machine learning model is introduced and the relevant knowledge of statistical theory is expounded,the basic principle of least squares support vector machine(LS-SVM)and the advantages of wavelet kernel function are analyzed,and LSWSVM model is constructed based on wavelet kernel.Secondly,according to the difficulty in choosing the parameters of LSWSVM model,the particle swarm optimization(PSO)algorithm is utilized to optimize the model parameters.In order to solve the problem that the basic PSO algorithm can easily converge locally,chaos variables are introduced to construct chaotic particle swarm optimization(CPSO)algorithm by means of the randomness and ergodicity of chaotic motion.Thirdly,for the defects that chaotic sequence is sensitive to initial value,a chaotic particle swarm optimization algorithm with extreme value mutation is utilized.The results show that the improved mutation chaos particle swarm optimization(MCPSO)algorithm has good performance in optimization by simulation analysis of classical test functions.The improved MCPSO algorithm is combined with LSWSVM prediction model.Finally,through the statistical analysis of the passenger flow data of Chengdu Metro Line 1,the MCPSO-LSWSVM forecasting model is constructed to further study the passenger flow of the daily station,the morning peak,the early rush hour and the sudden large passenger flow at the Sichuan Gymnasium Station,and mine the distribution and generation rules of passenger flow in urban rail transit.The prediction simulation of the different types of passenger traffic at the same time period is carried out,and the prediction results of the PSO-LSWSVM prediction model and the CPSO-LSWSVM prediction model is compared.The results show that the MCPSO-LSWSVM prediction model has higher accuracy and better generalization than the traditional prediction model,has more superiority in the prediction of passenger flow,and provides an effective decision-making method for the operation scheduling of urban rail transit in day-to-day operation and large-scale event of large passenger flow.