Devishetty100/clickbait-detector

Clickbait Detector

This model is a machine learning classifier trained to detect clickbait headlines. It uses a Random Forest algorithm with TF-IDF vectorization to classify news headlines as either "clickbait" or "real".

Model Details

Model Description

• Model type: Random Forest Classifier
• Task: Text Classification (Clickbait Detection)
• Input: News headlines (text strings)
• Output: Binary classification ("clickbait" or "real")
• Language(s) covered: English
• License: MIT

Model Sources

• Repository: Devishetty100/clickbait-detector
• Paper or resources: N/A
• Demo: N/A

Uses

Direct Use

This model can be used to classify news headlines and identify potentially misleading or sensationalized content. It can be integrated into content moderation systems, news aggregators, or educational tools to help users discern between genuine news and clickbait.

Downstream Use

• Content filtering and moderation
• Journalism education
• Social media analysis
• Research on media manipulation

Out-of-Scope Use

This model should not be used for:

• Automated content removal without human oversight
• Making decisions that affect individuals' livelihoods or rights
• Classifying content in languages other than English

Bias, Risks, and Limitations

Recommendations

Users should be aware that:

• The model may have biases based on the training data
• Performance may vary across different domains or writing styles
• False positives/negatives can occur
• The model is trained on English text only

Known Limitations

• Trained on a specific dataset which may not represent all types of clickbait or real news
• May not perform well on very short or very long headlines
• Does not consider context beyond the headline text itself
• Binary classification may not capture nuanced cases

Training Details

Training Data

The model was trained on the Clickbait Dataset from Kaggle, which contains news headlines labeled as clickbait or real.

• Dataset size: 32,000 samples (16,000 clickbait, 16,000 real)
• Data preprocessing: Text cleaning, TF-IDF vectorization with English stop words, max 5000 features
• Train/test split: 80/20 stratified split (25,600 train, 6,400 test)

Training Procedure

• Architecture: Random Forest with 200 estimators
• Hyperparameters: Default parameters except n_estimators=200, random_state=42
• Training time: [Not specified]
• Hardware: [Not specified]
• Software: scikit-learn, pandas, numpy

Evaluation

Metrics

The model achieves the following performance on the test set:

• Accuracy: 91.45%
• Precision: 0.92 (macro avg)
• Recall: 0.91 (macro avg)
• F1-Score: 0.91 (macro avg)

Testing Data, Factors & Metrics

Testing Data

• Same dataset as training, held-out test set
• Stratified sampling to maintain class balance

Factors

• Headline length and complexity
• Use of sensational language
• Topic domain

Metrics

• Accuracy, Precision, Recall, F1-Score
• Confusion Matrix

Results

              precision    recall  f1-score   support

   clickbait       0.89      0.95      0.92      3200
        real       0.94      0.88      0.91      3200

    accuracy                           0.91      6400
   macro avg       0.92      0.91      0.91      6400
weighted avg       0.92      0.91      0.91      6400

Environmental Impact

Estimated Emissions: Not calculated

Hardware Type: Standard CPU training

Hours used: [Not specified]

Technical Specifications

Model Architecture and Objective

• Architecture: Ensemble of decision trees (Random Forest)
• Objective: Binary classification using TF-IDF features
• Input preprocessing: TF-IDF vectorization
• Output postprocessing: Class prediction

Compute Infrastructure

• Hardware: CPU-based training
• Software: Python, scikit-learn

How to Use

Loading the Model

from huggingface_hub import hf_hub_download
import joblib

# Download model and vectorizer
model_path = hf_hub_download(repo_id="Devishetty100/clickbait-detector", filename="clickbait_detector.pkl")
vectorizer_path = hf_hub_download(repo_id="Devishetty100/clickbait-detector", filename="tfidf_vectorizer.pkl")

# Load
model = joblib.load(model_path)
vectorizer = joblib.load(vectorizer_path)

Making Predictions

# Example headline
headline = "You won't believe what happened next!"

# Transform and predict
features = vectorizer.transform([headline])
prediction = model.predict(features)[0]

print(f"Prediction: {prediction}")  # Output: 'clickbait' or 'real'

Requirements

• Python 3.6+
• scikit-learn
• joblib
• huggingface_hub

Citation

If you use this model, please cite:

@misc{clickbait-detector,
  title={Clickbait Detector},
  author={Devishetty100},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/Devishetty100/clickbait-detector}
}

Contact

For questions or issues, please open an issue on the repository.