Visualization with Chinese in XGBoost and RandomForest

When feature names or class labels contain Chinese characters, tree visualizations generated by XGBoost, scikit-learn, and Graphviz may render them as boxes or missing glyphs. The usual fix is to set a font that supports Chinese before Graphviz renders the image.

The examples below use FangSong. You can replace it with any Chinese-capable font installed on your system, such as SimHei, Microsoft YaHei, Noto Sans CJK SC, or Source Han Sans SC.

XGBoost

xgb.to_graphviz() returns a Graphviz object. Before rendering it, modify the Graphviz source and add a node font definition.

import re
import xgboost as xgb


def set_graph_font(graph, font_name="FangSong"):
    graph.source = re.sub(
        r"graph \[ rankdir=TB \]\n\n    0 ",
        f'graph [ rankdir=TB ]\n\n node [fontname="{font_name}" shape=plaintext]\n\n    0 ',
        graph.source,
    )
    return graph.source


diagraph = xgb.to_graphviz(model, num_trees=9)
diagraph.format = "png"
set_graph_font(diagraph)
diagraph

If the font still does not render, check the fonts available to Graphviz on the local machine. On Linux, fc-list is often the quickest way to verify whether the font is installed:

fc-list | grep -i "fang\|noto\|source han"

RandomForest

For scikit-learn’s RandomForestClassifier or RandomForestRegressor, export each tree to a DOT file, replace the default font with a Chinese-capable font, and then render the DOT file with Graphviz.

import re
from pathlib import Path
from subprocess import call

from sklearn.tree import export_graphviz


def plot_forest(model, column_names, output_dir="forest", font_name="FangSong"):
    output_path = Path(output_dir)
    output_path.mkdir(parents=True, exist_ok=True)

    for i, estimator in enumerate(model.estimators_[:10]):
        dot_path = output_path / f"tree-{i}.dot"
        cn_dot_path = output_path / f"tree-cn-{i}.dot"
        png_path = output_path / f"tree-cn-{i}.png"

        export_graphviz(
            estimator,
            out_file=str(dot_path),
            feature_names=column_names,
            class_names=["Class Name 1", "Class Name 2"],
            rounded=True,
            proportion=False,
            precision=2,
            filled=True,
        )

        source = dot_path.read_text(encoding="utf-8")
        source = re.sub(r"helvetica", font_name, source, flags=re.IGNORECASE)
        cn_dot_path.write_text(source, encoding="utf-8")

        # Convert to PNG using Graphviz.
        call(["dot", "-Tpng", str(cn_dot_path), "-o", str(png_path), "-Gdpi=600"])

In a Jupyter notebook, display one rendered tree with:

from IPython.display import Image

Image(filename="forest/tree-cn-0.png")

Before running the conversion step, make sure Graphviz is installed and available on the command line:

dot -V

If dot is missing, install Graphviz with the package manager for your operating system, then rerun the notebook or script.