Due to the curse of dimensionality, a reduction of the number of feature columns is sometimes required before you can get any work done. However, there are also other reasons for reducing the dimensions. For example, plotting and visualizing scatter data on a two-dimensional piece of paper or computer screen requires that you have only two dimensions to show.
There are two main strategies for reducing dimensions, as follows:
- Selection: Choose the best features and eliminate the others.
- Transformation: Create new features that summarize the combinations of the original ones.
