A Financial Fraud Behavior Prediction Method Based on Mobile Device Behavior Data

Abstract

The invention discloses a mobile device behavior data-based financial fraud behavior prediction method. The method comprises the following steps of screening out user behavior data from historical operation data of a target application; performing one-hot coding on the user behavior data and adding a corresponding timestamp; and obtaining a user behavior path feature; according to coding data andthe user behavior path feature, building a recurrent neural network model, and obtaining a fraud risk probability of a user. According to the method, the behavior data of the user in a mobile device is mined and modeled, behavior habits of the user are objectively and quantitatively mined, and a predicted probability of a fraud risk is given, so that the behavior data with the relatively low utilization rate previously is enabled to exert higher values, and an existing user risk assessment system is perfected.

Description

technical field [0001] The invention relates to the technical field of financial risk control, in particular to a method for predicting financial fraud based on mobile device behavior data. Background technique [0002] Consumer finance refers to the non-bank financial business that provides individual users with loans for consumption on the principle of small amounts and decentralization, approved by the China Banking Regulatory Commission. This type of business has the advantages of small single credit line, fast approval speed, no mortgage guarantee, and short loan period. [0003] China's consumer finance business has experienced rapid development in 2017. According to the statistics of Wangdaijia, as of October 2017, there are more than 2,000 related institutions, which have completed transactions of more than 200 billion yuan per month. Emerging consumption Financial products have well met the financial needs of the public. However, on the other hand, due to some bus...

AI Technical Summary

Problems solved by technology

[0006] 1) Behavior data collection scheme: Behavior data is large in size and has many types. Simple collection of all data without filtering will bring storage and calculation burdens and performance noise to further data mining. On the other hand, The collection of too much data will also increase the pressure of data transmission and affect the user experience

[0007] 2) Coding and feature engineering of behavioral data: The coding and feature engineering work of traditional methods when processing behavioral data needs to involve expert experience or a large amount of data analysis work. On the one hand, the cost of labor and time is quite high, on the other hand It is a commonly used feature engineering solution with complex logic and a large amount of code

However, user behavior habits will also be reflected in the order of page browsing and the order of operations on the page. Traditional methods are relatively limited for information mining of such fine behavior paths.

Images