Interactive Plot Example - Lightgbm

Load libraries

python

from pathlib import Path

import matplotlib.pyplot as plt
import pandas as pd

import lightgbm as lgb

%matplotlib inline

try:
    # To enable interactive mode you should install ipywidgets
    # https://github.com/jupyter-widgets/ipywidgets
    from ipywidgets import SelectMultiple, interact

    INTERACTIVE = True
except ImportError:
    INTERACTIVE = False

Load data

python

regression_example_dir = Path().absolute().parents[1] / "regression"
df_train = pd.read_csv(str(regression_example_dir / "regression.train"), header=None, sep="\t")
df_test = pd.read_csv(str(regression_example_dir / "regression.test"), header=None, sep="\t")

y_train = df_train[0]
y_test = df_test[0]
X_train = df_train.drop(0, axis=1)
X_test = df_test.drop(0, axis=1)

Create Dataset object for LightGBM

python

lgb_train = lgb.Dataset(
    X_train,
    y_train,
    feature_name=[f"f{i + 1}" for i in range(X_train.shape[-1])],
    categorical_feature=[21],
)
lgb_test = lgb.Dataset(X_test, y_test, reference=lgb_train)

Configuration dictionary

python

params = {"num_leaves": 5, "metric": ["l1", "l2"], "verbose": -1}

Training

python

evals_result = {}  # to record eval results for plotting
gbm = lgb.train(
    params,
    lgb_train,
    num_boost_round=100,
    valid_sets=[lgb_train, lgb_test],
    callbacks=[lgb.log_evaluation(10), lgb.record_evaluation(evals_result)],
)

Plot metrics recorded during training

python

def render_metric(metric_name):
    lgb.plot_metric(evals_result, metric=metric_name, figsize=(10, 5))
    plt.show()

python

if INTERACTIVE:
    # create widget to switch between metrics
    interact(render_metric, metric_name=params["metric"])
else:
    render_metric(params["metric"][0])

Plot feature importances

python

def render_plot_importance(importance_type, max_features=10, ignore_zero=True, precision=3):
    lgb.plot_importance(
        gbm,
        importance_type=importance_type,
        max_num_features=max_features,
        ignore_zero=ignore_zero,
        figsize=(12, 8),
        precision=precision,
    )
    plt.show()

python

if INTERACTIVE:
    # create widget for interactive feature importance plot
    interact(
        render_plot_importance,
        importance_type=["split", "gain"],
        max_features=(1, X_train.shape[-1]),
        precision=(0, 10),
    )
else:
    render_plot_importance(importance_type="split")

Plot split value histogram

python

def render_histogram(feature):
    lgb.plot_split_value_histogram(gbm, feature=feature, bins="auto", figsize=(10, 5))
    plt.show()

python

if INTERACTIVE:
    # create widget for interactive split value histogram
    interact(render_histogram, feature=gbm.feature_name())
else:
    render_histogram(feature="f26")

Plot trees

python

def render_tree(tree_index, show_info, precision=3):
    show_info = None if "None" in show_info else show_info
    return lgb.create_tree_digraph(gbm, tree_index=tree_index, show_info=show_info, precision=precision)

python

if INTERACTIVE:
    # create widget to switch between trees and control info in nodes
    interact(
        render_tree,
        tree_index=(0, gbm.num_trees() - 1),
        show_info=SelectMultiple(  # allow multiple values to be selected
            options=[
                "None",
                "split_gain",
                "internal_value",
                "internal_count",
                "internal_weight",
                "leaf_count",
                "leaf_weight",
                "data_percentage",
            ],
            value=["None"],
        ),
        precision=(0, 10),
    )
    tree = None
else:
    tree = render_tree(53, ["None"])
tree