Session Ready
Exercise

PCA - dimension reduction

In the previous exercise, you worked on a dataset with two variables. During an interview, you are likely to face a bigger dataset.

PCA allows reducing the number of variables without significant loss of informational value.

PCA returns the dataset of the same size as your original dataset. It's up to you, how many variables to keep!

The following parameters of prcomp() reduce dimensions based on:

  • tol - the standard deviation as percentage of the first component's standard deviation,
  • rank - the maximal number of components.

The letters dataset contains numerical attributes of letters.

Instructions 1/3
undefined XP
  • 1

    Identify the principal components and show the proportion of variance captured by each of them.

    • 2

      Omit components for which the standard deviation is lower than 0.25 of the first component's standard deviation.

    • 3

      Limit the number of principal components to 7.