Complaint early warning processing method and device

Abstract

The invention provides a complaint early warning processing method and device. The method comprises the following steps: obtaining complaint amount data in a preset first time period, screening abnormal time points from the data of each time point based on a statistics and rule method, labeling the abnormal time points, constructing sample data according to complaint amount data of normal time points and complaint amount data of abnormal time points labeled, and standardizing the sample data according to different complaint businesses and cities; for the data of each time point in the obtainedcomplaint amount data, features are constructed according to the accumulated complaint amount, and the constructed features are screened according to importance sorting; training the standardized sample data and the screened features to obtain a trained fusion model, and performing complaint early warning by using the fusion model; wherein the fusion model is combined with a logistic regression model and a random forest model. Automatic early warning of full-service incoming call complaints in the communication field can be realized, the fault discovery time delay can be shortened, and the early warning result is more accurate.

Description

technical field [0001] The embodiments of the present invention relate to the technical field of complaint handling in communication operations, and in particular to a complaint early warning processing method and device. Background technique [0002] The threshold setting rules of traditional early warning are basically "average value + b (b is a constant)" and "average value * c (c is a constant, usually a percentage)". This algorithmic rule based on historical experience may lead to the failure of early warning Hysteresis (the first is the growth rate of complaints when a problem occurs). In addition, the value of b needs manual adjustment, which consumes a lot of time. [0003] At present, there are some automatic threshold technologies. For example, in the prior art, a wireless communication network call complaint warning method is disclosed, which provides a wireless communication network that can accurately monitor the changes of mobile phone Internet access, VoLTE, ...

AI Technical Summary

Problems solved by technology

This method is based on the dynamic threshold technology of the clustering algorithm, and its core is the kmeans algorithm, but the kmeans algorithm is sensitive to noise and abnormal points, and the data imbalance leads to poor results.

[0004] The traditional and existing automatic threshold technology cannot adapt to the current changing business characteristics and complaint characteristics, and the discovery ability is limited

Images