Hands-On Predictive Analytics with Python

Dimensionality reduction using PCA

Dimensionality reduction techniques are procedures that take a number of features and generate a smaller number of features that seek to preserve the majority of the information from the original ones. Say, for example, that you have data about the socioeconomic dimensions of a household: income level, wealth, taxes paid, years of education of the mother, years of education of the father, size of the house, and so on. Many of these features will have a strong correlation because they reflect many aspects of the socioeconomic status of the household. Suppose that you have 20 such features; you can use dimensionality reduction techniques to create a set of, say, two features that will preserve, for example, ...