Practical Data Analysis Cookbook

Extracting the useful dimensions using Linear Discriminant Analysis

Now that we understand the mechanics (and trade-offs) of dimensionality reduction, let's use it for classification.

In this recipe, we will introduce Linear Discriminant Analysis (LDA). LDA, in contrast to the methods presented earlier in this chapter, aims at representing the dependent variable as a linear function of many other features; in that sense, it is similar to a regression (which we will discuss in the next chapter). The LDA shows similarities to the ANOVA analysis of variance and logistic regression in how it models the linear relationships in the data that capture (explain) the variance the best.

We will use a linear SVM classifier to test the effectiveness of our dimensionality ...