learning_rfc.py 文件源码

def feature_importance(rf, save_path=None):
    """Plots feature importance of sklearn RandomForest

    Parameters
    ----------
    rf : RandomForestClassifier
    save_path : str
    """
    importances = rf.feature_importances_
    nb = len(importances)
    tree_imp = [tree.feature_importances_ for tree in rf.estimators_]
    # print "Print feature importance of rf with %d trees." % len(tree_imp)
    std = np.std(tree_imp, axis=0) / np.sqrt(len(tree_imp))
    indices = np.argsort(importances)[::-1]
    # Print the feature ranking
    # print("Feature ranking:")
    # for f in range(nb):
    #     print("%d. feature %d (%f)" %
    #           (f + 1, indices[f], importances[indices[f]]))

    # Plot the feature importances of the forest
    pl.figure()
    pl.title("Feature importances")
    pl.bar(range(nb), importances[indices],
           color="r", yerr=std[indices], align="center")
    pl.xticks(range(nb), indices)
    pl.xlim([-1, nb])
    if save_path is not None:
        pl.savefig(save_path)
    pl.close()