Consumption credit scene fraud detection method based on ABC-SOM neural network

Abstract

The invention relates to the technical field of risk control of the Internet financial industry, in particular to a consumption credit scene fraud detection method based on an ABC-SOM neural network,wherein the adopted SOM network is a self-organizing mapping unsupervised neural network model without tutors, and the consumption credit scene fraud detection method can be applied to the Internet financial industry by automatically searching internal rules and attributes in data. Network parameters and structures are changed in a self-organizing and self-adapting mode, the method has nonlinearity, high parallelism, fault tolerance, robustness and high self-adapting learning capacity, and the capacity in the aspect of processing uncertainty or fuzzy information is outstanding. Compared with traditional optimization algorithms such as GA and PSO, the ABC algorithm is not prone to falling into local optimum, the convergence rate and the stability are both improved, the ABC algorithm is notlimited by constraint conditions of the ABC algorithm in a search space, and the method has the advantages of being good in robustness, few in parameter and the like. The SOM model optimized by adopting the ABC algorithm has the characteristics of high precision, strong reliability and the like, the accuracy of application behavior fraud detection can be improved, and the real-time detection of consumption credit scene fraud is realized.

Description

technical field [0001] The invention belongs to the field of risk control technology in the Internet financial industry, and specifically utilizes an artificial bee colony algorithm (Artificial Bee Colony Algorithm, ABC algorithm) optimized SOM neural network for fraud detection in consumer credit scenarios. Background technique [0002] Today's consumer credit is developing rapidly. In consumer credit anti-fraud, traditional detection methods mainly rely on prior knowledge based on pre-defined anti-fraud rules and supervised machine learning algorithms. The detected data level is usually the original attribute or It is fine-grained 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 o...

AI Technical Summary

Problems solved by technology

[0004] However, the performance of the SOM network combined with the BP neural network model is affected by weights and thresholds. In addition, the BP neural network uses the method of gradient descent to adjust the weights and thresholds, which makes it easy to fall into the defects of local minimum and slow convergence speed. Although it can be used Particle swarm optimization algorithm, etc., but it will increase the computational complexity of the algorithm, affect the training time and training effect of the model, and is not conducive to the real-time detection of fraudulent behavior in consumer credit scenarios. Therefore, to solve the above problems, a neural network based on ABC-SOM is proposed. Fraud Detection Method for Consumer Credit Scenarios

Images