Dimensionality Reduction in R

Course Description

Do you ever work with datasets with an overwhelming number of features? Do you need all those features? Which ones are the most important? In this course, you will learn dimensionality reduction techniques that will help you simplify your data and the models that you build with your data while maintaining the information in the original data and good predictive performance.

Why learn dimensionality reduction?

We live in the information age—an era of information overload. The art of extracting essential information from data is a marketable skill. Models train faster on reduced data. In production, smaller models mean faster response time. Perhaps most important, smaller data and models are often easier to understand. Dimensionality reduction is your Occam’s razor in data science.

What will you learn in this course?

The difference between feature selection and feature extraction! Using R, you will learn how to identify and remove features with low or redundant information, keeping the features with the most information. That’s feature selection. You will also learn how to extract combinations of features as condensed components that contain maximal information. That’s feature extraction!

But most importantly, using R’s new tidymodel package, you will use real-world data to build models with fewer features without sacrificing significant performance.

What You’ll Learn

Foundations of Dimensionality Reduction

Prepare to simplify large data sets! You will learn about information, how to assess feature importance, and practice identifying low-information features. By the end of the chapter, you will understand the difference between feature selection and feature extraction—the two approaches to dimensionality reduction.

Feature Selection for Model Performance

Chapter three introduces the difference between unsupervised and supervised feature selection approaches. You’ll review how to use tidymodels workflows to build models. Then, you’ll perform supervised feature selection using lasso regression and random forest models.

Feature Selection for Feature Importance

Learn how to identify information-rich and information-poor features missing value ratios, variance, and correlation. Then you’ll discover how to build tidymodel recipes to select features using these information indicators.

Feature Extraction and Model Performance

In this final chapter, you’ll gain a strong intuition of feature extraction by understanding how principal components extract and combine the most important information from different features. Then learn about and apply three types of feature extraction — principal component analysis (PCA), t-SNE, and UMAP. Discover how you can use these feature extraction methods as a preprocessing step in the tidymodels model-building process.