Machine Learning Based Predictive Modelling for Fraud Detection in Digital Banking

Banking stands out as a pioneering industry, having amassed vast amounts of structured data over its history and being among the earliest adopters of data science technologies. The world is witnessing rapid growth in banking sector and hence payment transaction method is changing into the use of cashless transaction instead of the use of cash money. Digital banking gives customers a better service quality in the areas of prepaid card, credit card, cashless payment, money transfer, etc. However, the security of digital banking has received attention due to the fraudulent behavior of fraudsters as the absence of sufficient digital banking security has kept many customers away from the service till today. Although fraud is not a new issue, activities related with fraud always cost the global economy billions of dollars each year. Fraudulent activities involve inclusively significant financial risks which may threaten profitability, and the image of the whole of an economic sector. This study aims to use machine learning based predictive modelling to implement to improve fraud detection and prevention, inside of a public or private economic entity. In this paper, this problem is focused on the propose an autonomous and effective intelligent new ensemble method for fraud transaction detection. Firstly, Hidden Markov Model (HMM) is used for observing the hidden states of financial transaction, that is based on the idea of the probabilistic models designed for sequences of observations where the underlying process is assumed to be a Markov process with hidden states. And then, machine learning model called Gradient Boosting Classifier (GBC) is applied to classify the fraud. Finally, our hybrid method integrates GBC and HMM. Experiments are conducted to ensure the effectiveness of HMM and GBC.