Credit Card Fraud Detection by Machine Learning with Feature Optimization Source Code

Project Discription

This model proposed a detection of credit card fraud  by use of genetic algorithm for feature optimization and Error Back prorogation neural network for learning.In this chapter gentic algorithm frog leaping were proposed.

Frog Leaping & EBPNN Explanation of proposed Frog Leaping algorithm for feature selection and training of neural network was done in this section of paper.  Training of neural network for credit card fraud detection. Whole work was divide into three module. Input training dataset pre-processing was done in first module. Feature selection wss done in second module by frog leaping algorithm. Training of neural network was done in third module of work.

Conclusion The development of effective systems for the identification of fraud is essential to the reduction of these losses, and an increasing number of these systems rely on methods of machine learning to aid fraud investigators. However, the design of fraud detection systems presents a particularly difficult challenge due to the non-stationary distribution of data, the highly unbalanced distribution of classes, and the lack of availability of labelled transactions as a result of confidentiality concerns. All of these factors combine to make the problem more difficult to solve. This work has developed a frog leaping based feature optimization algorithm, reduction in feature dimension improves the learning of the model. This work has further uses error back propogation neural network for training and prediction of credit card fraud. It was obtained that use of these genetic algorithm and machine learning model for credit card frad detection has increases the work efficicency. Expeirment was done on real dataset. Result shows that proposed model has improved the detection precision value by 0.9397% as compared to previous existing model. In previous model use of all feature for the training of model has increases the confusion, that’s leads to false class prediction. It was found the use of frog leaping based selected features with error back propogation has increases the accuracy parameter values by 10.007% as compared to previous existing method. 

Project Sample Image