Hybrid Machine Learning Approach for Data Anomaly Detection in Credit Card Transactions

Data anomaly detection is crucial for maintaining the integrity, security, and efficiency of systems, helping to detect and respond to abnormal events promptly. In this paper, a hybrid approach of a combination of two unsupervised machine learning algorithms, Isolation Forest (IF) and One Class Support Vector Machine (OCSVM) with one supervised machine learning algorithm Random Forest Classifier as an ensemble method is used in credit card transactions system for data anomaly detection. For this, credit card fraud detection dataset from Kaggle is used for anomaly detection. As credit card transactions data is huge, varying and unlabeled, unsupervised algorithms like IF and OCSVM become suitable for it. The ability of OCSVM of defining of what constitutes the anomaly, power of isolation forest to detect outliers efficiently and use of random forest classifier as an ensemble method of decision trees, complementing both IF and OCSVM are associated to increase accuracy of a system. The result of this paper shows that this hybrid approach provides more accuracy than individual algorithm.