Optimizing Machine Learning Data Pre-Processing for Financial Fraud Detection

Financial fraud is an expensive issue that can be difficult to detect and machine learning (ML) algorithms are frequently applied. Most datasets for financial fraud are imbalanced with few fraudulent and many non-fraudulent transactions. Our research focuses how data-balancing techniques effect the accuracy of three common ML models when applied to financial fraud detection. We experimented with Resample, SMOTE, ADASYN, and RandomOverSampler for data balancing then apply K-Nearest Neighbours (KNN), Support Vector Machine (SVM), and Random Forest (RF) algorithms to see how well they perform in the detection of financial fraud after balancing.