Machine Learning-Based Algorithmic Approach for Enhanced Anomaly Detection in Financial Transactions

Governments all over the world aim to encourage digital money transaction to minimize administration cost, avoid customer's physical present while purchasing, avoid money loss from theft, and to save time. So nowadays, people have started to use cashless transaction. Due to the increasing nature of cashless transaction, credit card becomes common nowadays, and many people have adapted to this payment mode. Customers are expanding on a regular basis as a result of the flexibility afforded by credit card transactions. Because of their ubiquity, credit card payments will not come without any risk. There are many types of credit card frauds, for instance, skimming, phishing, card lost or stolen, etc. and by using some machine learning algorithms, many fraud detection systems are available in recent years but the efficiency of those algorithms is not appreciable. Genetic algorithm, k-means algorithm, artificial neural network, logistic regression, SVM, etc. are few algorithms used to develop the system to detect credit card fraud. This paper makes use of an algorithm called isolation forest, which helps to detect the outlier data points in a better way. It performs well in large dataset, and also, it provides better results and accuracy when compared to SVM and local outlier factor.