Published Paper Details:

According to studies of international monetary data, the sum of money lost due to fraud in all transactions around the world keeps rising. In particular, the adoption of the Bank-as-a-Service model by financial institutions will place additional strain on payment processing infrastructure and exacerbate the already serious problem of payment fraud. Using automated machine training and Big Data analysis methods, this research aims to synthesis viable solutions for detection of fraud in electronic payment systems. Online Payment fraud has been given a lot of attention since it can result in significant economic losses and negative consequences for account holders. In order to construct a reliable model for identifying fraud, it is essential to employ efficient feature engineering. It is discovered, however, that the present feature engineering that relies on transaction frequency has flaws. Convenient as they may be, fraudsters are drawn to modern money transfer services to conduct scams in which unsuspecting victims are tricked into sending money to fake accounts. As a result of the limitations of conventional rule-based methods for detecting fraud, machine learning models have found widespread application. Most studies build features by extracting patterns using raw transaction records because learning directly from transaction data is challenging owing to its enormous dimensionality. Recency, frequency, financial, and anomaly detection are some of the common labels used to classify these characteristics in the existing research. Using real-world transaction data, features can be extracted and most relevant features can be selected to train the machine learning model for accurate fraud detection. The temporal properties of user transactions are shown using frequency-based feature engineering, however the characteristics of fraud and the differentiation of transaction behaviors are not taken into sufficient account. In this research a Multi Level Associated Weighted Feature Vector (ML-AWFV) Model using machine learning is proposed for accurate feature extraction and selection for further processing of online payment fraud detection. The proposed model when contrasted with the existing