Machine Learning for Detecting Credit Card Frauds

Abstract

Credit card frauds has been a threat that has evolved as a major source of loss for the financial sectors. It has been seen in the different parts of world causing loss of billions of dollars. It is also a area which needs attention from the researchers as the task of fraud detection can be automated using the different machine learning classifiers and data science. If the frauds model encounter the fraudulent transactions it will raise an alarm to the system administrator. The paper proposes a model which uses the machine learning classifiers to detect the fraudulent transactions. The classifiers used in the paper are SVM (Support Vectore Machine ), Isolation Forest and Local Outlier. The focus of the research is to detect the fraudulent transactions to 100% and also we emphasise on the fact that no normal transaction should be detected as fraud wrongly. The process starts with preprocessing the data and then the classifers are applied. The results from each classifers is evaluated to check the one with the better performance. The performance can be increased with use of deep learning algorithms but with the rise in expennses.