Communication user consumption trend detection method based on clustering algorithm

Abstract

The invention is applied to the field of service provider individual client consumption trend identification, and provides a cluster analysis identification algorithm based on consumption time sequence data. The cluster analysis identification algorithm is used for communication consumption. The algorithm comprises steps of establishing a consumption sequence of historical consumption data of a client, carrying out clustering calculation of a consumption time sequential sequence of the client, and determining a client consumption trend according to a time sequence clustering result. The algorithm can effectively resist abnormal consumption interference data, distinguish starting and ending time of different consumption periods and different consumption levels, and provide an accurate data basis for client consumption trend analysis. The algorithm can be used for consumption trend identification of individuals and groups at other aspects such as income losses, consumption abnormity, consumption periods and the like.

Description

technical field [0001] This utility model patent involves the analysis of the characteristics of the consumption trend of users in the communication industry; Background technique [0002] The analysis of time series is widely used in scientific research, engineering applications, and even social and economic fields; time series analysis modeling is based on certain data assumptions, such as obeying a stationary process ARMA model, or a stationary process ARIMA model after differential processing; but , these model assumptions are difficult to fully apply to the analysis of the actual telecom customer consumption sequence; generally speaking, the customer's communication consumption should be a stable process, but due to competitors, alternative services, and other uncertain and abrupt factors Different degrees of impact on customer communication consumption; [0003] We extracted a set of customer consumption sequences, the following is the classification of general user c...

AI Technical Summary

Problems solved by technology

[0007] Communication user consumption does not conform to the "Gaussian independent and identical distribution" (I.I.D) assumption. The main reason is that although customer consumption has a certain degree of randomness, most consumption sequences show a certain continuity, which is consistent with the previous period of consumption. Correlation is not a complete "noise" distribution; the identification method based on time sliding window is not ideal, it is difficult to determine the appropriate window size, and the statistical results show great randomness, which affects normal business analysis; if simply "time " and "monthly consumption" two-dimensional vectors for general clustering processing, such as using the K-means algorithm, there may be cluster segmentation in a non-strict sequence, and the clusters that are segmented cross each other in time, which is not conducive to timing analysis;

Images