Method and device for automatically obtaining suspected card raising channel

Abstract

The invention discloses a method and device for automatically obtaining a suspected card raising channel, and relates to the field of intelligent risk control. The method comprises the steps of collecting basic features of a to-be-identified user; acquiring input basic characteristics of a user to be identified, performing weighted fusion through a trained random forest model with acquired suspected card raising rules in combination with a trained linear regression model to acquire a suspected card raising user list in the user to be identified, and acquiring a corresponding card raising channel according to the suspected card raising user list. A random forest learning base algorithm fused by a plurality of decision tree classifiers is applied to extract potential suspected card raising rules from sample data, weighted fusion is carried out on the potential suspected card raising rules and a linear regression classification model, the integration generalization ability is enhanced, and the suspected card raising rules have diversity. Supervision and management of a card raising phenomenon in a daily operation process of a market department are satisfied so that a channel violationbehavior is effectively controlled.

Description

technical field [0001] The invention relates to the field of intelligent risk control, in particular to a method and device for automatically obtaining suspected card maintenance channels. Background technique [0002] In recent years, the development of the information and communication industry is entering a new era. The penetration rate of communication users is getting higher and higher, and the trend of homogeneity is becoming more and more obvious. The development of traditional telecom service users has also encountered growth bottlenecks, competition is becoming increasingly fierce, and low-value users are increasing. . At the same time, the remuneration paid by operators to social channels is on the rise, and KPI orientation has led to malicious violations in some channels, that is, through some technical means, illegally raising cards to obtain agency fees or related remuneration from mobile companies. According to statistics, in January 2017, Zhejiang Mobile’s su...

AI Technical Summary

Problems solved by technology

[0004] The existing scheme is to complete the learning task by building multiple learners through ensemble learning, that is, to use multiple weak classifiers to form a strong classifier, while the traditional method is to use a weak learner such as a decision tree, which has poor generalization ability. Poor, and it is greatly affected by the disturbance of data samples, so it is impossible to deeply dig out suspected card users, and the risk control effect is poor

Images