Embedded methods integrate feature selection directly into the learning algorithm. This technique embeds the selection process within the training of a predictive model. You know filter and wrapper techniques are discrete processes.
- Combines the benefits of filter and wrapper methods
- Train only one machine learning model for feature selection
- Computational cost is not severe since it is embedded within the training
Methods available
- Decision tree or Random Forest Classifier
- Lasso (L1 regularization)
- Ridge(L2 regularization)