Random Machine Learning Pipeline with Feature Importance Table Generation

This function creates a randomly chosen machine learning pipeline using CountVectorizer for text vectorization and LogisticRegression or DecisionTreeClassifier as the classifier. Then, it uses the show_weights function from the eli5 library to generate a feature importances table and formats it as an HTML table.

Technology Stack : eli5, scikit-learn, CountVectorizer, LogisticRegression, DecisionTreeClassifier

Code Type : Machine learning

Code Difficulty : Intermediate

                
                    
import random
import eli5
from eli5.formatters import table
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.pipeline import Pipeline

def random_eli5_function(texts, labels):
    # Create a random pipeline with a CountVectorizer and a classifier
    pipeline = Pipeline([
        ('vect', CountVectorizer()),
        ('clf', random.choice([eli5.models.LogisticRegression(), eli5.models.DecisionTreeClassifier()])),
    ])
    
    # Fit the pipeline on the provided texts and labels
    pipeline.fit(texts, labels)
    
    # Generate a feature importances table
    feature_importances = eli5.show_weights(pipeline.named_steps['clf'], vec=pipeline.named_steps['vect'])
    
    # Print the feature importances in a formatted table
    print(table.format_as_html(feature_importances))

# Sample usage
texts = ["This is a sample text.", "Another sample text.", "Yet another text sample."]
labels = [0, 1, 0]

# Call the function
random_eli5_function(texts, labels)