CREDIT CARD FRAUD DETECTION USING ANOMALY DETECTION IN UNSUPERVISED MACHINE LEARNING
The unauthorized use of a credit or debit card, or rather a card number, to frequently acquire money or possessions is a recurring criminal offence. Thus, detection and elimination of such nefarious activities can be beneficial towards individuals as well as financial institutions. This paper proposes a simple, yet effective method to spot suspicious transactions among several valid transactions by implementing several methods of anomaly detection. The system is programmed on Python, through the means of Jupyter notebook. Local outlier factor; to quantify anomaly scores will be used in conjunction with the isolation force algorithm. The aforementioned algorithms will comb through the dataset of over 280,000 transactions and predict fraudulent transactions. As intended, the system would be able to detect abnormal transaction patterns with high accuracy and thereby prove to be a valuable asset of exposing criminal minds.