Machine Learning for Fraud Detection in Banking Systems

Money transfers have significantly increased as a result of the quick development of technological advances, and fraud has also increased in tandem. The overall declines resulting from forged payments worldwide are continuously increasing, according to data on worldwide economic data. The digital transformation of financial interactions will make payment fraud worse by placing more strain on payment providers. Consequently, identifying fraudulent activity has emerged as an intriguing subject. Employing real-world unbalanced databases this study applied a machine learning (ML) oriented system for banking data fraud detection. To fix the problem of the groups not being balanced, the Synthetized Minority over-sampling technology (SMOTE) was employed to resample the database. It was found that the Adaptable Booster (AdaBoost) method worked best with these machine learning techniques when it came to classification. To test the methods, performance matrices were used. The tests showed that using AdaBoost makes the suggested methods work better. In addition, the improved processes led to better results than the old methods. The research's findings support the efficacy of creating fraud identification methods for e-commerce platforms utilizing automated ML techniques. Through adopting the results of the research into practice, banks may be able to lower the time and money required to create and update active platforms against fraudulent payments and increase the efficiency of money transaction surveillance.