The Method Of Resident Travel Demand Prediction And Its Application Based On PSO-LSSVM Machine Learning

As a complex social ecosystem,cities gather high-density populations and socioeconomic activities.As the urbanization process continues to accelerate,residents' transportation needs are diverse,and the number of motor vehicles is growing rapidly.The existing transportation services are difficult In line with the city's development level,road traffic congestion,air pollution,and frequent traffic accidents have caused traffic difficulties,which have greatly affected the normal operation of city functions.How to solve transportation problems has become the key to urban development.Restricted and influenced by technology,space and economy,the method of dealing with transportation problems has been transformed into construction and transportation demand management by focusing on transportation infrastructure investment.The traffic demand forecasting method,which is the basis of traffic demand management,shifts from the classic lumped method to the research of the non-aggregated method,and the Logit model is typical of the non-aggregated method.With the rise of machine learning,it is also possible to predict the traffic travel of residents,which provides a new direction for the application of nonaggregate models and the assessment of short-term traffic demand management and control strategies.This paper establishes a PSO-optimized LS-SVM machine learning method to predict residents' travel needs,and realizes the prediction of residents' travel modes,travel times and time.The main content of this paper:(1)Through the travel survey data of residents,the travel characteristics are analyzed from three main aspects: travel mode,travel time and travel times,grasp the differences of travel of residents,and obtain the trends and laws of travel of residents.(2)According to the residents' personal attributes,family attributes and travel attributes of the proportion distribution relationship and overall characteristics,and divide and quantify each attribute and travel behavior choice,and use Spearman correlation coefficient to analyze the correlation of influencing attributes,Identify influencing factors that are significantly related to travel behavior.(3)Based on the PSO-LSSVM machine learning method,a travel mode prediction model,a trip number prediction model,a resident's first travel time prediction model and a last travel time prediction model are constructed.(4)Perform 50% cross-validation on the travel mode prediction model to verify the stability and generalization ability of the model.In order to compare the prediction effects,the same sample data is used to make predictions with standard SVM and MNL models,respectively,and the prediction accuracy is compared and analyzed.(5)As an application of the model,analyze the impact of the development and optimization of public transport resources on residents' travel behavior,and conduct a single-factor control variable to predict and analyze the residents' travel behavior from the residential house to the nearest bus stop walking time and waiting time,Based on the analysis results,put forward countermeasures and measures for the priority development of public transportation and optimal allocation of resources for the urban traffic planning and traffic provide a scientific basis for the establishment of related development policy.Figure 63,table 51,110 references.