ibps-so 2016 · Question 131 of 394

Two exercises are proposed upon the same customer le. In the rst, every one of forty thousand past personal-loan accounts is already labelled in the data as having turned non-performing or not, and a model is to be built from those labels which will place each fresh applicant into one of the two classes. In the second, no label of any kind exists: the customers are simply to be divided into a number of groups such that the customers within a group resemble one another in their balances and their transaction behaviour more than they resemble the customers of any other group, the groups themselves being discovered by the algorithm and not given to it. The two exercises are respectively instances of:

1. AClustering, which is supervised learning; and classi cation, which is unsupervised learning
2. BClassi cation and regression, both of them supervised, the di erence lying only in the number of classes
3. CAssociation rule mining and sequential pattern mining, both of them unsupervised
4. DBoth are unsupervised, since in neither case is the analyst able to tell the algorithm which answer he expects
5. EClassi cation, which is supervised learning because the class labels are known in advance and the model is trained upon them; and clustering, which is unsupervised learning because no labels are supplied and the groups are discovered from the data itselfCorrect