Chapter 9. Fraud Analytics Using Autoencoders and Anomaly Detection
Detecting and preventing fraud in financial companies, such as banks, insurance companies, and credit unions, is an important task in order to see a business grow. So far, in the previous chapter, we have seen how to use classical supervised machine learning models; now it's time to use other, unsupervised learning algorithms, such as autoencoders.
In this chapter, we will use a dataset having more than 284,807 instances of credit card use and for each transaction, where only 0.172% transactions are fraudulent. So, this is highly imbalanced data. And hence it would make sense to use autoencoders to pre-train a classification model and apply an anomaly detection technique to predict possible fraudulent transactions; that is, we expect our fraud cases to be anomalies within the whole dataset.
In summary, we will learn the following topics through this end-to-end project:
- Outlier and anomaly detection using outliers
- Using autoencoders...
