The strategies for this are divided into 2 broad types: fraud detection as well as consumer activity analysis. The initial category of strategies works with controlled recognition processes at transaction stage. Transactions are classified as illegitimate or regular depending on preceding historical evidence in such systems. This dataset can then be used to construct classified models that can forecast the status of new documents (normal or fraudulent). A standard two-classification function, including rule inference, decision trees, as well as neural networks, has various model development approaches. This method has been shown to accurately identify most previously found fraud techniques, often known as identification of misuse essential to illustrate the main discrepancies in an overview of consumer behaviour and methods to fraud investigation. The system of fraud detection can identify established tricks from fraud, with a small false positive rate. Such schemes derive the sign as well as pattern of fraudulent strategies provided in the revelation data set as well as can therefore quickly decide precisely that frauds; the machine is witnessing at the moment.