Credit Card Fraud Detection Using Anomaly Techniques

Abstract

Credit card fraud transaction detection system is a method used for determining the fraudulent transactions that take place every once in a while. The project uses a test data set of around 27,000 credit card transactions which have been taken from Caltech (Kaggle). The project comprises of primarily 2 major algorithms and uses anomaly detection as a method to classify the fraudulent transactions. The Local Outlier Factor (LoF) defines the various parameters that have to be used in determining the criteria for fraudulent transactions. It then checks upon the different transactions for the various parameters present in the given LoF. This factor then gives each transaction a score based on the various transactions that have or will have taken place. These scores can range from 0 - 1. Each transaction is thus given a score which is based on the various parameters given in the LoF. The second part of the project is isolation forest algorithms which is an algorithm that isolates the transaction which have a high rate of anomaly detected in them. Thus, these transactions are isolated and then checked with various parameters to be labelled as either fraudulent or real transactions. The algorithm also uses charts to check for spikes in the average transaction. We also uses technique like data visualization in order to show the output in more understandable ways which may include histograms, graphs and matrix .Through these two algorithms and with help of data visualization technique we can detect the fraudulent transactions from correct transaction and obtain results in quick time, Since these algorithms are much more time efficient than other machine learning algorithms in this type of tasks.