International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Since from the last few years there is a significant increase in credit card transactions are playing a vital role. Thus, it is leading to significant financial losses everywhere in present days. It is very challenging task to process the huge amount of data and it is making data sets unbalanced and complex. There are basically two major problems while handling data. It is analysed with fraud and non-fraud transactions, and it doesn’t contain relevant, appropriate, and correlated data that affects their prediction performance in a negative way. Followed by, it has involved the interest of machine learning (ML), which consists of fraud detection as a main theme. It has been involved by various ML methods such as Logistic Regression (LR), Support vector machines (SVM), Decision Trees (DT), Random Forest (RF), and K-Nearest Neighbours (KNN). However, the above methods cannot meet the excellent performance required to find and predict abnormal fraud patterns. In this project the main contribution is to provide a framework for fraud detection (FFD). Firstly, we have to overcome the unbalanced data issue, the framework uses an under sampling technique. Followed by, we have to select the relevant features by applying the feature selection (FS) mechanism. Next, Neural networks is mainly builds the ML model and it aims to handle the capability, a modified version of the Particle Swarm Optimization (PSO) algorithm, Polynomial Self Learning PSO (PSLPSO), is proposed for hyper parameters C and σ. Finally, the framework’s effectiveness is depicted in the experimental results on a transaction dataset of real credit card.