The Regression Analysis Of TD User’s Satisfaction With Network Index

With the competitive between telecom operators increasingly, improve customersatisfaction for the operator to maintain the stock of users gain more potential users hasgreat significance. Operators can learn the difference of service between competitorsthrough satisfaction surveys. It’s can provide a strong reference for enhancing services.China Mobile TD network users to improve customer satisfaction survey networkingdirection, positioning a profound impact on the significance of each factor.In this study, the background extraction TD network user satisfaction surveys targetcustomers district-level indicators related to TD network, e.g. turning, dropped calls,switching and other indicators. Using the statistical software R language, the above indexand TD user network satisfaction survey results regression analysis. Trying to buildrelationship between satisfaction and background data network.It will providing a referenceguide to enhance satisfaction TD user network from a network perspective.Choose from variable regression model and the focus of this paper is to solve. In caseof a small amount of data, much amount of variables, and multicollinearity relactionshipbetween variables, the model results prove that classical multivariate linear regressionanalysis has limitations. By trying different methods of regression analysis modle and theindependent variable options: ridge regression model fitting parameter estimation accuracyupgrade; LASSO regression approach to enhance the accuracy and variable selection;selected with the idea of principal component regression Partial Least squares, principalcomponent regression analysis substitution variables to solve the problem ofmulticollinearity between the independent variables. This article uses the TD network usersatisfaction and network back-indicator data, use the above method to establish a model,and finally select a optimal model. The results show that, although ridge regression andLASSO regression method can improve the accuracy of the model to a certain extent, thepurpose of variable selection, but unable to solve the problem of multicollinearity totally,making the model established perception and reality discrepancies. The partial least squaresmethod based on principal component regression ideas is a good solution to the problem ofmulticollinearity.Article ultimately affect TD find satisfaction in the customer’s network through the keyfactor partial least squares method: PS domain wireless connection TD network, thesystem switches between the packet domain network success rate (cell) indicator.Regression model, the use of the new season, with key indicators resident district survey predict user satisfaction with the results better.