Hybrid proactive approach to prediction of credit card fraudulent activities detection

• Main Authors: Prabha, N., Manimekalai, S.
• Format: Conference Proceeding
• Language: English
• Subjects: Algorithms; Datasets; Deep learning; Fraud; Machine learning; Performance evaluation
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0178874

Financial crime’s negative prediction is important research and it is needed for financial institutions. Previous research using single and hybrid algorithms has been used to detect credit card fraud. Because no further research into different hybrid algorithms for a given dataset was done, these approaches have significant limits. This study creates and tests hybrid machine learning models for detecting fraudulent behaviour using a real-world dataset. The different states of the survey described different machine learning related algorithms that are used to detect fraud and also hybrid algorithm mechanisms are also used to produce better performance in terms of accuracy and related activities. The prediction is made based on the proactive and reactive types. Most of the previous types of work are performed based on the reactive type. This work proposed a proactive type prediction using different dynamic activities such as types of traction, time, location, customer amount, transaction ranges, etc. In this work, a proactive approach to prediction of fraudulent activity detection using ensemble learning and deep learning techniques is proposed. The eXtreme Gradient boosting ensemble method is used for proactive feature selection and CNN deep learning methods are used for analysis of fraud activities. For implementation, a real-time dataset from a commercial bank with 284, 807 features is used. The accuracy, sensitivity, and specificity metrics are used to measure the performance of proposed work.