Question

Explain whether each scenario below is a regression, classification or unsupervised learning problem. If it is a supervised learning scenario, indicate whether we are more interested in inference or prediction. Finally, provide in each case the number of observations, n, and the number of predictors, p.

(a) Oil excavation is a very expensive process and the oil resources are not distributed uniformly in an area, so it is important to find the best spots for oil extraction. To do this engineers consider a very coarse grid (each edge length is on the order of miles), dig a well at the vertices of the grid and take a sample of the sand there. 24 different measurements are taken from each sand sample. An engineer has sand samples for 35 locations where they know the results of the digging (how much oil was present at that location). Additionally, the engineer has sand samples for 80 prospective well locations, and wishes to find the most promising spot to dig a future well.

(b) An online retailer must decide whether to display advertisement A or advertisement B to each customer on the basis of collected customer demographics (age, zip code, and gender). A set of 300 of its customers have already expressed a preference for one advertisement or the other.

Answer 1

(a) Regression problem. Here, we are more interested in prediction because we need to detrmine which among the prospective locations are having high chances of being oil-rich. No. of observations n = 35; No. of predictors = 24.(Because we measure 24 different features of the collected sand sample. These features are predictors.)

(b) Classification problem. Here ,we are more interested in prediction because the retailer wants to decide whether to display ad A or B to each customer. No. of observations n = 300; no. of predictors = 3 (age, zip code, gender).