Fraud detection via deep neural variational autoencoder oblique random forest

Fraud detection is critical problem of many financial company that has been researched both in academy organizes and industry. The objective of this paper is fraud detection for transaction level in credit card and e-commerce. The new machine learning model for fraud detection is proposed that deep neural variational autoencoder oblique random forest. Variational autoencoder is strong by regarding the distribution of latent variables and by the most optimal connecting weights between latent variables and visual variables are successful by maximizing log likelihood of observed variables. The proposed method is combined the advantage of variational autoencoder and oblique random forest for fraud detection problem. The proposed method is applied for two data sets being credit card data set and e-commerce data set. The experimental results show that the proposed method attains the better performance than the single or the other methods which used the same data sets.