ABSTRACT
This chapter focuses on illustrating each algorithm's domain of applicability, their relative strengths and limitations, trade-offs, and any alternative methods. It examines linear regression for its applicability and also because it serves to illustrate the fundamental principles of interpretability and explainability. The chapter discusses recommender systems as they combine elements of supervised and unsupervised learning, and are neither but a concept on another level. Predictive analytics cannot truly be separated from descriptive analytics because without a good understanding of the working dataset it is often not possible to build a robust predictive model. A key ingredient for the success of semi-supervised learning is being able to label the unlabelled data as well as possible. Linear regression is a technique that models the relationship between a response variable and one or more explanatory variables. A key assumption that is made in linear regression besides assuming that a linear relationship exists between the variables, is that the data points are independent.
