Research On Forecast Passenger Flow Bursted In Urban Rail Station On Data Mining

Among the past years,the development of area of informationalization has forced the technology of data mining to be a hot field all over the world.As for the area of transportation,data mining has been applied to provide data source for several of models and arithmetic.Nowadays,to mine and recognize the abnormal passenger flow of urban rail is one of important branches,especially for the abnormal flow caused by large organizations.This is because that those mass flow,which are discrete processes,sudden and aperiodicity would have a huge impact on the operation in subway station.There is no doubt that if the sudden mass flow could be predicted ahead,as a result,station organizer can make better strategy to cope depending on the forecast result.How to help relative departments make more scientific measures to deal with mass flow on the basis of providing the accurate forecast consequence is one research topic with realistic significance,especially on the condition that a large amount of mega-events are hold every year and 2022 XXIV Olympic Winter Games will hosted in Beijing.Although more and more researchers have applied the data mining on dealing with and analyzing the data about transportation,there is less people who mainly focus on improving speed of mining and reducing computation complexity to combine the traditional transportation and new technology.Aiming at these shortages,SAX(Symbolic Aggregate approximation)is firstly introduced to reduce dimensions of massive time series of passenger flow and discovering association rules on analyzing characteristics of urban rail passenger flow at common day and abnormal stage.As a result,the difficulty to cope with data is successfully reduced.And for the purpose of guaranteeing that the fundamental information are distortionless and meanwhile calculation counts can be the lowest when mining the flow data,the original measurement to compute similarity between two time series,Euclidean Distance,is transformed to DTW(Dynamic Time Warping),and that is optimized in constraint of searching path in this paper.The research result shows that the optimized arithmetic can assure to be both on improving mining efficiency and recall rate.In addition,improving the defect on depending on single entering flow to forecast the short term passenger flow is as the research goal,and the forecast model combining the entering and exiting passenger flow caused by activities is construed in paper.What is more,the thesis discusses the use of Statistics and Granger test of causality to demonstrate that the exiting flow of activities can be contributed to forecast the entering flow.Besides,as for forecast model,initial parameters of wavelet neural network are optimized by genetic algorithm.And the model is trained and learnt from sample data discovered by improved method.On conclusion,the forecast error on the double-factor time series is obviously better than the single-factor time series.