Detecting Financial Fraud Using Anomaly Detection Techniques: A Comparative Study of Machine Learning Algorithms

Financial fraud presents a substantial challenge to the American economy, culminating in substantial monetary losses and breaching the integrity of financial systems. The focal objective of this research paper was to resolve the prevalent issue of financial fraud detection in the USA by performing a comparative study of multiple machine learning algorithms, particularly concentrating on their anomaly detection capabilities. Experimentation was performed using various machine learning classifiers, such as logistic regression, random forest, Multi-layer perceptron, SVM, Naive Bayes, AdaBoost, decision tree, and KNN. Data utilized for this study was retrieved from Kaggle's website (https://www.kaggle.com/mlg-ulb/creditcardfraud). The five-metrics used for the performance evaluation were accuracy, precision, recall, F1-score, and confusion matrix. Decision Tree had superior performance at classification accuracy, followed closely by AdaBoost, then KNN and Random Forest, as per the outcomes obtained in this study. Implementing the proposed models has an array of benefits to both financial organizations and the US economy in terms of real-time fraud detection, advanced accuracy of fraud detection, cost efficiency, reduction in financial losses as well as strengthening financial organizations.