Several Machine Learning algorithms can be used for classifying payment transactions as fraud or genuine one. The algorithms used for example are Logistic Regression, Random Forest, Naive Bayes and Multilayer Perceptron. Each algorithm can be used for fraud detection with high accuracy. ML model can be used for detection of other irregularities.

Unsupervised learning such as clustering can be used for detecting fraud. you should finding outlier with having a great clustering on your data.

K-means algorithm can be beneficial for these problems.