Statistical Analysis And Prediction Of User Complaints In Internet Tourisl Enterprises

In recent years,with the development of the Internet industry,especially the explosive development of tourism in recent years,consumer users have become the main driving force for the progress of enterprises,and the interests of enterprises are closely related to users.However,with the diversification of business and the increasing number of users,the number of complaints received by enterprises in the tourism industry has also increased sharply every year.In order to ensure that the enterprise's own profits are not damaged,major operators should actively analyze and study the user's behavior,and take appropriate countermeasures to users before they complain,so as to effectively reduce the probability of users' complaints.This study begins with the observation of the modern Internet consumer market where user complaints frequently occur.User data comes from the Internet tourism enterprise,where to go,which is now the leading brand of domestic tourism search APP.Through the statistical analysis of the relevant characteristics of registered users,we can accurately predict the occurrence probability of their complaints in the future.This paper mainly makes statistical analysis of user's own characteristics and behavioral-characteristics,feature engineering,which depicts user portraits in detail.Finally,user characteristics are divided into five categories:user's basic characteristics,special group characteristics,historical singleton characteristics,other statistical features and user's behavioral characteristics,and new uses are added as far as possible in the combination of various features and factors affecting user complaints.Household characteristics,improve user portraits.Next,the paper establishes a mathematical model to predict the probability of user complaints.Two XGBoost models and two LightGBM models are selected,and the four models are fused,and the results of the four models are calculated logically.The result is the probability of user complaints predicted by the model.Finally,the article chooses the appropriate threshold for the model to screen out the users who are susceptible to complaints,and adopt reasonable coping strategies for them,so as to achieve the result of reducing the rate of users' complaints.In the feature engineering part,185 user features are finally obtained.These features are generated by combining the factors affecting user complaints,but not all of them are effective.Finally,the speed and efficiency of operation are comprehensively considered in the model.Sixty features of higher importance are selected to replace all features in the model training.Although the effect of the model is slightly reduced,the execution efficiency is improved highly.In this paper,four machine learning models are designed and implemented.Finally,a complete user complaint model is obtained to predict the user's complaint probability.After many training and screening,the accuracy and recall rate of the optimal model are as high as 70%.After the project is launched,the effect is better.In the past month,the company's user complaint rate has been reduced from about 10%to less than 3%.It can basically accurately identify more than 90%of user complaints.In this paper,there are some innovations in research direction and research methods.Choosing the direction of user complaints is to see the problems existing in the process of enterprise operation,and make a reasonable and accurate prediction of user complaints by using more comprehensive and appropriate features and models.The research results of this paper have strong practical application.For enterprises,effective analysis of user behavior characteristics and prediction of the occurrence probability of user complaints are one of the important ways to improve enterprises and promote enterprises.In addition,formulating reasonable improvement measures to face this series of problems is a two-way thinking between managers and users,which is not only conducive to enhance the user product experience,but also conducive to enterprises fully understand their own shortcomings and make timely changes.The research and analysis of the user complaint prediction project can give relevant enterprises reasonable thinking,the occurrence of user complaint behavior is not traceless,and the improvement of corporate reputation benefits is not impossible.