permutation_feat_importance.py

def feat_permutation_importance(df, feats, model):
    """
    Takes in a dataframe of 'plays', features list
    and model, plots and returns the mean score across
    all the folds
    Parameters
    ---------
    df : a dataframe object
            Contains the plays
    feats : a list object
            Contains the features' columns
    model : a string object
            The model name

    Returns
    -------
    importances : pd.Series object
            Contains the importances of features
    """

    # define the dataset features and target
    X = df[feats]
    y = df["gm_cluster"]

    # initialize Random Forest model
    model = model

    model.fit(X, y)

    # perform permutation importance
    results = permutation_importance(model, X, y, scoring='f1_weighted')

    # get importance
    importance = results.importances_mean
    idxs = np.argsort(importance)
    importances = pd.Series(importance, index=feats)

    # plot feature importance
    plt.title('Permutation Feature Importance', fontsize=12)
    plt.barh(range(len(idxs)), importances[idxs], align='center')
    plt.yticks(range(len(idxs)), [feats[i] for i in idxs])
    plt.xlabel('Feature Importance')
    plt.show()

    return importances

# check for all the features
feat_permutation_importance(train_norm_df, train_norm_feats, logres)

# re-define norm_feats
train_norm_feats.remove('START_POSITION_n')

# check for all the features != 'START_POSITION'
feat_permutation_importance(train_norm_df, train_norm_feats, logres)