Credit card companies must monitor for fraudulent transactions in order to keep their customers from being charged for items they have not purchased. Such data is unique in being extremely imbalanced, with the particular dataset we will be working on in this chapter having fraud constituting 0.172% of the total transactions. It contains only numeric input variables, which are the result of a PCA transformation, and the features Time and Amount. The Time feature contains the seconds elapsed between each transaction and the first transaction in the dataset. The Amount feature is the amount transaction, a feature that we will use, for instance, in cost-sensitive learning. The Class feature is the response parameter and, in case of fraud, it takes the value 1, and 0 otherwise.
So what is example-dependent, cost-senstive learning? Consider the costs associated...
