Research On Subway Complaint Text Mining And Satisfaction Evaluation Based On LDA Model

With the rapid development of society and the continuous progress of urban rail transit technology,urban rail transit has become one of the important ways for people to travel in Beijing,and has an important impact on the daily life of the capital citizens.The operation service quality is one of the important indicators to measure the operation level of urban rail transit,and the general public has higher requirements for the operation service of urban rail transit nowadays.How to deal with passengers' complaints quickly and effectively,improve the quality of rail transit service,and improve passenger satisfaction,has become the most concerned thing of scholars and managers in this field.Passengers' complaints are often presented in the form of text with a large volume.It takes time and effort to analyze the text manually.Therefore,it is necessary to apply natural language processing technology and algorithm to realize the theme mining,automatic classification and complaint content analysis of subway complaint text in order to discover the deficiencies in the operation service timely and effectively,and improve the service quality of Beijing Subway.The main research contents include the following aspects:(1)analyzing the Beijing subway ride service process and service characteristics,dividing the subway ride process into 9 links,and decomposing the characteristics of each service link,clearly defining the personnel and related equipment that provide the service are undertook,and then texting the service characteristics analysis;analyzing passengers' psychological needs for subway service complaints,complaint channels and the necessity of handling subway complaints.(2)The LDA topic modeling method is introduced,and the topic of the subway complaint text is excavated through the LDA theme modeling algorithm.Combined with the subway service process and service feature analysis in Chapter 3,the subway service complaint text is divided into 8 topic types,and an automatic text classifier is built to realize the automatic classification of complaint texts.(3)Through calculating the frequency of keywords co-occurrence,the keyword co-occurrence network of complaint text is constructed and the visualization analysis and association analysis of keyword co-occurrence network are realized.And through the form of case analysis,the complaint cases of temperature control noise,station car service and safety,which are the most complained in the complaint text,are analyzed in depth,so as to tap the real needs of passengers and provide a basis for Beijing Subway.(4)This paper studies the index setting,sampling method,sample structure and index calculation method of Beijing subway satisfaction survey,and puts forward improvement suggestions for the existing passenger satisfaction evaluation through comparative analysis of the satisfaction survey results and the research results of complaint text mining.