Analysis of Variant Data Mining Methods for Depiction of Fraud

Abstract

Financial fraud is a growing problem with far-reaching concerns in the financial sector. Online transaction is the basic problem that raises many fraudulent quires around the world which cause loss of money to the people. These transactions generated huge volume of complex data in daily life. The depiction of fraud from credit card is still a key challenge due to two main reasons: firstly, profiles of ordinary and fraudulent behavior changes with the Passage of time, and secondly highly skewed credit card fraud records. Therefore, this study considered this challenge and proposed the solution to identify the fraudulent transactions through the credit cards using data mining techniques. Data mining has played a significant role in identifying credit card fraud from online transactions. Dataset collected from the publically available source and refine it. The employed classifiers are Naive Bayes, Bayes net, Logistic regression, Random forest, Decision tree, support vector machine, Decision stump, K- Nearest Neighbor, J48 and Binary Classification Technique. These techniques are applied on the preprocessed data. This data consists of 284,785 credit card transactions. Extensive experiments were conducted. The accuracy of each classifier was recorded in order to perform comparison. Our empirical analysis spotlights that K-NN outperforms in term of accuracy which is 99.95% than other classifiers. The findings of this study would be useful for the banking sector.