Combined Prediction Model For Travel Time Considering Uncertainties And Its Applications

Accurate prediction information of travel time is an important indicator for traffic participants to make travel decisions,and is the basic theory of the implementation of intelligent transportation systems.Most of the previous individual prediction models for travel time ignored the uncertainties of the problem itself and the model selection.Combined prediction model can reduce the influence of uncertainty factors to a certain extent,but there are still such problems in determining the weight coefficients of component predictors of the combined prediction model,which will cause the deviation of weight coefficients calculation and further deteriorate the combined prediction performance of travel time.Focus on the above mentioned uncertainty factors affecting the combined prediction of travel time,this paper proposes a new combined prediction model for travel time based on fuzzy soft set with the help of the advantages of soft set and fuzzy soft set theory in dealing with uncertain problems.The main contents of this paper are as follows:(1)Analysis of main factors affecting travel time and selection of component predictors of combined forecasting.In reality,there are many factors that affect the travel time of vehicles,and in most cases,the travel time may be affected by multiple factors at the same time.This paper extracts the relevant information of multiple influencing factors based on the card-swiping data of expressway toll stations and the supporting traffic incident data,and then analyzes the main influencing factors of travel time based on fuzzy soft set theory.Meanwhile,considering the influence of subjective factors from the modeler,there will be a certain degree of uncertainty and randomness in the selection of component predictors,which will affect the combined prediction of travel time.Therefore,in order to ensure the rationality of the selection of component predictors,this paper evaluates the forecasting performance of some individual travel time prediction models based on fuzzy soft set theory,and choose the ones with better prediction performance as the final components of the combined prediction model for travel time according to the evaluation results.(2)Determination of the weight coefficients of component predictors and the combined prediction model for travel time.In addition to the uncertainty and randomness in the selection of component predictors of the combined prediction model,the prediction model itself is also uncertain.For example,when the same prediction model is used to perform the forecasting of different travel time series,its prediction accuracy will also be different.In order to further reduce the influence of uncertainties on the combined prediction results of travel time,this paper proposes methods for determining the weight coefficients of component predictors based on fuzzy soft set theory,and establishes the corresponding combined prediction model for travel time.Finally,the validity and superiority of the proposed model are verified through experiments and comparative analysis.(3)Construction of travel time function based on fuzzy soft set and its application in traffic assignment.Construct the travel time function based on the aforementioned methods and ideas,and then a novel travel time function which can flexibly consider multiple factors is presented.And the corresponding user equilibrium traffic assignment model is also proposed.The simulation analysis results has proved the validity of the proposed travel time function and its superiority in traffic assignment is also verified.