Supervised learning

In Chapter 3, Building a Fraud Detection System, we built a fraud detection model using clustering, which is an unsupervised learning method.

In this chapter, we can see a difference between supervised and unsupervised machine more clearly. In the case of credit card approval data, we have a clearly defined target variable, the result. We are only interested in the value of this target variable for new customers. The historical card approval data may be captured from a manual or automated card approval process.

You should keep in mind when a supervised learning method predicts the value of the target variable, it can never be 100% certain about the outcome. The process of building the model produces a model which provides a probabilistic estimate for the outcome for the data which has not been seen previously or the future data.

Supervised learning has two distinct steps. In the first step, we build the model by mining the historical data. Once the model is ready, it can...