Analysis And Research On Customer Complaints Of Shanghai Liquefied Gas Company Based On Text Classification

In 2018,Shanghai Gas company was established to guarantee the safety service and supply of liquefied gas cylinders in Shanghai.In order to manage the whole process of LPG cylinder service well and adapt to the requirements of future information development,the company has developed and used the customer information management system platform of LPG cylinder at the early stage of establishment.The management of the company uses this system management platform to record the text content of the complaints of LPG cylinder users and deal with the complaints through manual classification.Over time,the number of users and orders increased.The number of complaints accumulated in the system platform also gradually increased.How to realize automatic classification and scientific,efficient and accurate analysis of users’ complaint data has become an urgent problem to be solved.The development of natural language processing technology and big data algorithm provide a new direction for traditional manual classification processing.In this paper,text classification algorithm and word co-occurrence network analysis method are used to automatically classify and analyze the complaint text of LPG cylinder users.This is conducive to improve the accuracy and scientific nature of the company’s handling of complaints from LPG cylinder users,and reduce labor costs.This can improve customer satisfaction,reduce customer churn and facilitate the marketing of LPG cylinders.In this paper,the actual users of Shanghai Liquefied Gas Company as the research object,based on the user complaints and suggestions as the data,using natural language processing algorithm and big data algorithm to analyze the user complaints.Firstly,the LPG industry corpus is constructed,and its data comes from the company’s LPG business documents and service documents,policy management documents issued by government departments,LPG field websites and open industry literature.Then,the LPG industry dictionary is constructed by the degree of freedom and degree of condensation algorithm.Combining with the LPG industry dictionary established in the previous step,the text classification model is constructed by combining n-gram language model and naive Bayes algorithm to realize the automatic classification function of users’ complaints and opinions.Finally,keywords are extracted from different types of complaint text,and the co-occurrence network graph of complaint text words is constructed to realize the visual display and association analysis of different types of complaint problems.The cooccurrence network diagram can be used to analyze the core keywords of complaints of different types of users,further analyze the causes of complaints,and determine the core problems of complaints of this category.Through network analysis method,it is convenient for LPG company managers to put forward accurate rectification schemes for key problems of complaints and improve customer satisfaction.The comparison of experimental results shows that the classification accuracy of user complaints based on the constructed professional LPG industry dictionary is higher than the traditional text classification method directly.The validity and feasibility of this classification method are verified.LPG company can formulate corresponding service solutions and design reasonable,effective and scientific management schemes according to the analysis results of customer complaints.To comprehensively improve the company’s business management and service level,ensure supply capacity,further improve user service experience and operation management efficiency,improve the brand image of LPG company,and achieve smart gas service provider.