User location preference extraction algorithm based on exponential regression and maximum likelihood estimation

Abstract

The invention relates to the field of telecommunication service support, and provides an algorithm for calculating user behavior preference on the basis of exponential regression and maximum likelihood estimation. The algorithm comprises the steps that distribution analysis on user position data and abnormal data removal are conducted; the linear relation between the user activity frequency and the user quality state attribute is fit through negative binomial regression, parameters to be estimated are solved through maximum likelihood estimation, whether parameter estimation passes an inspection or not is checked, and then a regression equation is obtained. Three-quarter quantile of the expense, the flow and call duration is solved, and a critical point of the base station activity frequency is determined through the regression equation, so that the high quality state is achieved while the activity frequency of screened users is guaranteed. According to the algorithm, the problem that a traditional preference extraction algorithm is inaccurate and prone to be disturbed is solved, the preference behavior of the user can be comprehensively taken into account, and the marketing success rate can be increased while marketing recommendation is conducted on the user.

Description

Technical field [0001] The invention relates to the field of telecommunication service support, in particular to a user location preference extraction algorithm. Background technique [0002] In the actual operation of telecommunications, there is a wealth of active information about user base stations, but it is difficult to determine whether users are really active or not from simple active times. If you can correlate some quality indicators of users, such as consumption, calls, traffic, etc., analyze the internal correlation between the number of active times and these indicators, and use these indicators to help define the critical point of whether the user is active. At present, the user's location preference simply analyzes the number of times the user's base station location is active, looks at its distribution, and takes a higher score as the critical value for whether the user is active. There are too many subjective factors, lack of scientific data support, and at the ...

AI Technical Summary

Problems solved by technology

There are too many human-subjective factors, lack of scientific data support, and at the same time, it is impossible to know the quality of users, so it is difficult to guarantee the pertinence and accuracy of marketing

Images