Fraud Detection System for Financial System Using Machine Learning Techniques: A Review

Using machine learning approaches, this study presents a comparative investigation of fraud detection in financial systems. The study examines a number of databases, machine learning algorithms, and pre-processing methods that are employed in fraud detection. A variety of preprocessing techniques, feature extraction, and ML and DL approaches—including logistic regression, decision trees, random forests, K-NN, CatBoost, SVM, isolation forests, CNN, and RNN—as well as exploratory data analysis techniques are included in the literature review. The difficulties with feature engineering, class disparity, and machine learning technologies are all covered in the article. The UCI Machine Learning Repository, GitHub, Kaggle, and other publications and datasets are among the sources of information that the study uses to presentand analyse its results. The data sets include a variety of fraud situations such as application fraud, data phishing, online and offline fraud, credit card theft, and data theft. The study comes to the conclusion that machine learning algorithms offer a viable way to identify fraud, and that machine learning models improve over time and become more skilled at identifying fraud as new patterns appear.