Banking fraud is defined as any dishonest act or behavior to obtain privileged banking information without authorization from the user for monetary gain.
Retaining profitable customers takes the top priority for most banks and credit card fraud is a significant threat to this goal. The rise in digital payment platforms is leading to an exorbitant rise in fraudulent transactions that threaten the sanctity of these banks and the trust of customers. Thereby, credit card fraud detection using machine learning is not just a trend but a necessity to proactively monitor and deter its occurrence.
The dataset used in this study contains credit card transactions for September 2013 by several European cardholders. The dataset, however, is heavily unbalanced; only 0.172% of all transactions are classified as fraud and its features had been transformed using PCA to protect classified information, leaving only amount and time fields as it is.
The objective of the study is to create a model using the given dataset that best identifies fraudulent transactions.
Steps:
- Data pre-processing and Exploratory Data Analysis
- Model Building
- Model Evaluation