Financial fraud is one of the major causes of monetary loss in banks and financial organizations. Rule-based fraud-detection systems are not capable of detecting advanced persistent threats. Such threats find ways to circumnavigate rule-based systems. Old signature-based methods establish in advance any fraudulent transactions such as loan default prediction, credit card fraud, cheque kiting or empty ATM envelope deposits.
In this chapter, we will see how machine learning can capture fraudulent transactions. We will cover the following major topics:
- Machine learning to detect fraud
- Imbalanced data
- Handling data imbalances
- Detecting credit card fraud
- Using logistic regression to detect fraud
- Analyzing the best approaches to detect fraud
- Hyperparameter tuning to get the best model results
