As you may have noticed, features are scaled in the previous section before training machine learning algorithms. Feature transformations are usually necessary for ML algorithms to work properly. For example, as a rule of thumb, for ML algorithms that use regularization, normalization is usually applied to features.
The following is a list of use cases where you should transform your features to prepare your dataset to be ready for ML algorithms:
- SVM expects its inputs to be in the standard range. You should normalize your variables before feeding them into the algorithm.
- Principal Component Analysis (PCA) helps you to project your features to another space based on variance maximization. You can then select the components cover most of the variance in your dataset, leaving the rest out to reduce dimensionality. When you are working with PCA...