Deep Learning for Real Credit Card Data

In this study, two major applications are introduced to develop advanced deep learning methods for credit card data analysis. Credit card information is contained in two data sets; credit approval dataset and card transaction dataset. The credit card dataset has two problems. One problem is using credit card approval dataset, it is necessary to combine multiple models, each referring to a different clustered group of users. The other problem is using card transaction dataset, since the actual unauthorized credit card use is very small, these imprecise solutions do not allow the appropriate detection of fraud. To solve these problems, we proposed deep learning algorithm to apply credit card dataset. The proposed methods are validated using benchmark experiments with other machine learnings. To evaluate our proposed method, we use real credit card transaction dataset from real system. The proposed methods are validated using large scale transaction dataset. Deep learning parameter adjustment is difficult. By optimizing the parameters, it is possible to increase the learning accuracy. The proposed method was able to confirm high discrimination accuracy even with less training data. Since fraud detection accuracy for each dataset is not stable, over fitting for datasets needs to be avoided. In addition, it is necessary to consider the calculation cost of deep learning.