ABSTRACT
Anomaly detection techniques are a well-known area for solving a particular kind of problem when there is the presence of an anomaly. Anomaly is the deviation from standard or we can say normal (regular) pattern. Anomalies are the data points that do not match the expected pattern. Anomalies can occur in transactions like Credit Card Fraud, Campaign Response. It also can occur in Video Surveillance like in Traffic, and can also as Spam, Malware/Intrusion. As technology is evolving day by day, it gives us many profits but at the same time many losses are also incurred. Like with the evolution of the domain of e-commerce and online banking, there is a drastic increase in the usage of credit cards for payment. Mostly the shopping (online shopping) is done by making the use of credit cards and due to this big usage of the credit cards; we are also facing the problem of fraud and big damages. So, it is very important to detect fraud in transactions so that we do not have to face big financial losses. There are many Machine learning algorithms or techniques, both supervised and unsupervised are available for anomaly detection. Here we can classify the fraudulent transaction in credit card fraud as point anomaly, so for detecting it we can use supervised algorithms like SVM, KNN, Random forest algorithm, and unsupervised algorithms like Isolation Forest algorithm, LOF algorithm. In this paper, we are discussing our experiment on credit card fraud detection by using Unsupervised Anomaly Detection algorithm. Later, we will also compare the performance of all the algorithms that we are using (Isolation forest Algorithm and LOF Algorithm).
