Multilabel Classification

In sklearn, Multilabel Classification assigns multiple labels to a single instance, allowing models to predict multiple outputs simultaneously. This method differs from traditional classification, where each instance belongs to only one class.

Scikit-learn offers tools like OneVsRestClassifier, ClassifierChain, and MultiOutputClassifier to handle multilabel classification and enable efficient model training and evaluation.

Syntax

Here’s the syntax for using multiabel classification in sklearn:

from sklearn.multioutput import MultiOutputClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split

# Step 1: Initialize the base classifier
base_model = RandomForestClassifier(random_state=42)

# Step 2: Create a MultiOutputClassifier wrapper for multilabel classification
multi_label_model = MultiOutputClassifier(base_model)

# Step 3: Train the model using the training dataset
multi_label_model.fit(X_train, y_train)

# Step 4: Make predictions on the test dataset
predicted_labels = multi_label_model.predict(X_test)

# Step 5: Evaluate predictions or use the results
print(predicted_labels)

• RandomForestClassifier: The base classifier for multilabel classification.
• MultiOutputClassifier: A wrapper to extend the base classifier for multilabel tasks.
• Training and testing: The model is trained with fit() and predictions are made using predict().

Example

This code demonstrates multilabel classification using scikit-learn by training a model to assign multiple labels:

from sklearn.datasets import make_multilabel_classification
from sklearn.ensemble import RandomForestClassifier
from sklearn.multioutput import MultiOutputClassifier
from sklearn.metrics import classification_report

# Generate synthetic multilabel data
X, y = make_multilabel_classification(n_samples=100, n_features=10, n_classes=3, n_labels=2, random_state=42)

# Initialize a base classifier
base_classifier = RandomForestClassifier()

# Wrap the base classifier for multilabel classification
model = MultiOutputClassifier(base_classifier)

# Train the model
model.fit(X, y)

# Predict labels for new data
predictions = model.predict(X[:5])

# Display predictions
print("Predicted Labels for First 5 Samples:")
print(predictions)

The code results the following output:

Predicted Labels for First 5 Samples:
[[1 1 0]
 [1 1 0]
 [0 0 1]
 [1 1 1]
 [0 1 0]]

Codebyte Example

The following codebyte example trains a Random Forest classifier for multilabel classification on dataset and predicts multiple categories for new samples:

us
Visit us
code
Hide code
Code
Output
Hide output
Hide output
Loading...
Run