Constructing a Model to Predict Fraudulent Credit Card Transactions using Machine Learning Techniques
As there is an increase in electronic payments volume, the economic stress on the fake detection of credit cards is becoming a fundamental challenge for service providers and financial institution, accordingly continuous forcing is done for enhancing the systems of fraud detection. As the e-commerce is so popular now a day, the credit card fraud is also becoming evenmore serious. Hence, the analysis on the detection of fraud is important and attractive. In view of financial services, the fraud in credit card is a critical problem because of every year there is a loss of Billion dollars. In the research studies there is a deficiency on reviewing the data of credit card owing in the real world to the problems of confidentiality. There are 2 major aspects due to the scarcity of practical concern. First one is the way and timing based on the supervised information available and the second is the measures utilized for the evaluation of the performance of fraud detection. In this paper, we suggest a methodology to construct a model which predicts fraudulent credit card transactions by making use of Machine Learning Techniques. We have applied single and ensemble classifiers to build the fraud detection model. We have developed different models by applying various Supervised Learning algorithms. And we have compared the performance of these models using Precision and Recall measure asthe accuracy is not correct measure for the imbalanced data. The class imbalance problem has been overcome by using SMOTE analysis. We have shown that the model build using Random Forest has given better performance than the other classifiers. So we conclude that this model can be used to predict credit card fraudulent transactions.