Abstract

Fraud has for decades been a major problem for merchants, especially for the online business sector that deals with credit cards.The challenging problem of fraud detection is that fraudsters make all possible efforts to make their transaction more legitimate. Another difficulty is that the number of legitimate records is far greater than the number of fraudulent cases. Such unbalanced sets require additional precautions from the data analyst. An effective technique for accurate fraud detection lies in developing dynamic systems that evolve to new fraud patterns. This implies that fraud detection must evolve continuously, and much faster than fraudsters which necessitate the hybridization method used in this research to tackle credit card fraudulent activities’ detection. A credit card fraud dataset used was obtained from the Kaggle machine repository. The result achieved showed that the Random Forest Classifiers as a machine learning proffers a significant performance for data split of 75:25 training to testing distribution with an astonishing result percent of 98% than the Artificial Neural Network as a deep learning which depicts an accuracy score of 0.9184 value, which is equivalent to 92%. This revealed the viability of the hybridized model used in this research for detection of credit card fraudulent activities within the record of a financial transaction with higher percentage accuracy compare to other researches.