Optimizing Regression Predictors with RFECV Random Feature Elimination with Cross Validation

Optimizing Regression Predictors with RFECV Random Feature Elimination with Cross Validation RFECV starts with all features, and then recursively eliminates the least relevant features until a specified number of features is reached. RFECV performs Random Feature Elimination with corss-validations. To find parameters, we use: RFECV.n_features_: The optimal number of features. RFECV.support_: A Boolean array of selected fatures. This makes it easy to filter by only the selected features. features = df[:, RFECTV.support_] y_predicted = RFECV.predict(X): Use model to predict values. Cross Valdiation: we train a model on one subset, and then test on another subset. We often use K-cross-fold validation. We partition the data into k equal subsets, train the model on k-1 subset, and then test on the remaining subset. We repeat this k times. This helps to reduce the likelihood of overfit. We can provide RFECV with a scoring model. We provide this in the scoring attribute. scikit-learn.org lists all of the possible scoring models. We need to scale the data to properly preform an RFECV analysis. We can use StandardScaler for this.