基于Python的Wine数据集决策树构建及可视化问题求助

Hey there, let’s work through this decision tree visualization issue together— I’ve dealt with plenty of similar hiccups using the Wine dataset and scikit-learn, so here are the most common fixes and checks:

Common Issues & Solutions #

1. Missing or Outdated Dependencies #

Your show_tree function relies on a few packages that are easy to miss, plus some deprecated methods. Let’s fix that first:

• Install all required packages if you haven’t already:

  pip install scikit-learn pydotplus graphviz pillow
  

• The misc.imread method from scikit-image is no longer supported. Replace it with Pillow’s Image.open() instead. You’ll need to import Pillow at the top of your script:

  from PIL import Image
  

  Then update the line in your function to:

  img = Image.open(path)
  

2. File Path Permissions or Invalid Path #

Double-check that the path 'dec_tree_01.png' is valid and you have write access to that location. If you’re using a relative path, it might be saving somewhere unexpected—try using an absolute path to eliminate confusion, like:

show_tree(dt, features, '/users/your_name/documents/dec_tree_01.png')

Also, if you’re saving to a subdirectory that doesn’t exist yet, add code to create it first (I’ll include this in the revised function below).

3. Mismatched Tree or Features Parameter #

Make sure:

• dt is a properly trained DecisionTreeClassifier instance. Confirm you initialized your classifier before fitting:

  from sklearn.tree import DecisionTreeClassifier
  c = DecisionTreeClassifier()
  dt = c.fit(X_train, y_train)
  

• The features list has exactly the same number of elements as the number of features in X_train. A mismatch here will throw an error in export_graphviz.

4. System-Level Graphviz Installation #

Even if you installed the graphviz Python package, you might need the system-level Graphviz binaries to generate the PNG:

• Ubuntu/Debian:

  sudo apt-get install graphviz
  

• macOS (Homebrew):

  brew install graphviz
  

• Windows: Download the Graphviz installer, run it, and add the bin folder to your system PATH.

Revised Working Function #

Here’s an updated version of your show_tree function that fixes deprecated methods, adds directory creation, and cleans up visualization:

import io
from sklearn.tree import export_graphviz
import pydotplus
from PIL import Image
import matplotlib.pyplot as plt
import os

def show_tree(tree, features, path):
    # Create target directory if it doesn't exist
    dir_path = os.path.dirname(path)
    if dir_path and not os.path.exists(dir_path):
        os.makedirs(dir_path)
    
    # Export decision tree to DOT format
    dot_data = io.StringIO()
    export_graphviz(
        tree,
        out_file=dot_data,
        feature_names=features,
        filled=True,  # Add color to nodes
        rounded=True,  # Rounded node edges
        special_characters=True
    )
    
    # Generate and save PNG image
    graph = pydotplus.graph_from_dot_data(dot_data.getvalue())
    graph.write_png(path)
    
    # Display the tree
    img = Image.open(path)
    plt.figure(figsize=(20, 20))
    plt.imshow(img)
    plt.axis('off')  # Hide axes for a cleaner view
    plt.show()

Give this a try, and if you’re still running into a specific error message, share it and we can dive deeper!

内容的提问来源于stack exchange，提问作者Jack