Internet financial platform application fraud behavior detection method based on fuzzy C-mean value

Abstract

The invention discloses an Internet financial platform fraud behavior detection method based on a fuzzy C-mean algorithm, and the method comprises the steps: carrying out the Z-score normalization anddimensionality reduction standard processing of collected information obtaining real-time measurement point data during the registration of a client account of an Internet platform, dividing the datainto a training set and a verification set, initializing the parameters of a fuzzy C-mean, and adopting a fuzzy clustering effectiveness function to automatically optimize the initial clustering number, obtaining a fuzzy C-mean clustering model through a target function, determining a classification decision rule according to the training set, classifying the verification set, and optimizing themodel with the application behavior and post-loan performance of the user; and deploying the optimized fuzzy C-mean value model to the rear end of an internet financial platform to perform online anomaly detection and monitoring on application behaviors of customers, sending out system early warning for applications in suspected abnormal states, and performing manual approval links or rejecting the applications. The method is advantaged in that high early warning result accuracy and strong fraud identification capability are achieved, and financial fraud risks are reduced.

Description

technical field [0001] The invention belongs to the technical field of risk control in the Internet financial platform industry, and specifically uses a fuzzy C-mean algorithm to provide a method for detecting whether there is fraud in an Internet financial platform application. Background technique [0002] In the anti-fraud of the Internet financial platform, the traditional anti-fraud detection method is mainly based on pre-defined anti-fraud rules and supervised machine learning algorithms based on prior knowledge. The detected data level is usually the original attribute or fine-grained level data. In today's big data era, financial risk dimensions are usually hundreds or thousands and extremely complex, making it difficult to formulate effective anti-fraud rules from a single or a few attributes, and supervised machine learning needs to accumulate a large number of performance samples for training. Models cannot identify new types of fraud in a timely manner. In resp...

AI Technical Summary

Problems solved by technology

[0004] However, there are two defects in the application of the traditional fuzzy C-means algorithm: first, the value of the initial clustering number c can only be selected artificially based on experience, and cannot automatically optimize the initial clustering number; Points and noise data are sensitive and easy to fall into local optimum, resulting in deviation in classification, which in turn affects the accuracy and stability of abnormal detection

Images