Understanding when to normalize data

You just learned that although data normalization is not always needed, it rarely hurts. Identifying features with different ranges is a good indication that you might need to apply data normalization, unless the Machine Learning model you are planning to build can handle this type of features.

You plan to cluster FIFA players based on their shot power and release clause amount (in millions EUR). Your clustering algorithm is sensitive to features having dissimilar ranges, so you wonder whether data normalization is required. In this exercise, you will first calculate the range of these two features using a sample of 100 FIFA players. Then you will produce a scatter plot of these features. The dataset is available as fifa_sample and the ggplot2 package is already loaded.

Question

Which of the following scenarios does NOT require data normalization?

Possible Answers

Clustering buyers according to their spending patterns via K-means.
Reducing the number of features in a microarray gene expression dataset through Principal Component Analysis.
Building a regression tree to predict the net value of a FIFA player.
Predicting whether a patient will be readmitted to the hospital in the next 6 months or not using K-nearest neighbors.

Exercise

Understanding when to normalize data

Instructions 1/4

Question

Possible Answers