Study And Application Of Non-parametric Regression Method And Bayesian Network

With the development of society,some new problems have emerged in the fields of economy,social sciences,transportation,etc.These problems are often characterized by high degree of non-linearity and uncertainty.Non-parametric regression methods can just describe nonlinear characteristics of the problem,therefore,it is concerned by researchers.Bayesian network is one of the most effective theoretical models in the field of uncertain knowledge representation and reasoning.In recent years,it has been used in data mining,fault analysis and other fields.In this paper,the related theories of non-parametric regression and Bayesian networks are studied,and their limitations in application are improved.They are applied to the prediction of short-term passenger flow in rail transit and the diagnosis of the cause of the large passenger flow.The main contents are as follows:1.The concept and principle of non-parametric regression are discussed.Four nonparametric regression methods are introduced.The non-parametric local regression is selected as the key research object,and the four methods of non-parametric local regression are introduced and compared in detail.Then,the K-nearest neighbor non-parametric regression is selected as the research target.The components and influencing factors of K-nearest neighbor regression are analyzed in detail.A non-parametric regression method based on error feedback is proposed for the problem of no feedback.2.Introduce the basic theoretical knowledge of Bayesian network and its representation and composition.The prior probability or conditional probability of Bayesian network node events is uncertain,and the use of expert knowledge will lead to subjective errors.The fuzzy set theory is introduced to quantify the probability of nodes.While determining the probability of nodes,the subjective errors are reduced as much as possible.Finally,the general application steps of Bayesian networks based on fuzzy set theory are given.3.Introduce the non-parametric regression method into the field of rail transit passenger flow prediction.The passenger flow statistics characteristics of rail transit are analyzed in detail.In order to ensure the matching data has practical significance,a classification sample database is constructed based on this,Based on the passenger flow data of Shanghai Metro Line 1 from April 1 to April 30,2015,the non-parametric regression method was used to predict short-term passenger flow and compared with other forecasting methods.4.Analyze the causes of the large passenger flow in the rail transit.Based on the relevant data of large passenger flow in Shanghai Metro Line 1 from 2014 to 2016,the Bayesian network for the cause of large passenger flow in rail transit is constructed.The fuzzy set theory is used to quantify the conditional probability of partial node events in Bayesian network,and uses the constructed Bayesian network to infer the probability of large passenger flow,and diagnoses the main cause of large passenger flow after the occurrence of large passenger flow.