AI Models Dashboard
Transparency and performance metrics for the AI models powering our news verification tool.
Model Ensemble Approach
Our news verification system uses an ensemble of multiple models to achieve higher accuracy and reduce bias. Each model contributes differently based on the type of content being analyzed.
Overall Accuracy
Combined model performance
87.6%
Based on our evaluation dataset of 10,000 news articles
False Positive Rate
Legitimate news marked as misleading
3.2%
We prioritize minimizing false accusations
Detection Rate
Misleading content correctly identified
91.4%
Across various types of misleading content
Model Comparison
Performance across different metrics
Naive Bayes Model
A probabilistic classifier based on applying Bayes' theorem with strong independence assumptions between features.
Performance Metrics
Accuracy84.2%
Precision82.7%
Recall87.5%
F1 Score85%
Strengths
- Fast training and prediction
- Works well with high-dimensional data
- Performs well with text classification tasks
- Requires less training data
Limitations
- Assumes feature independence (often not true)
- Less accurate with numerical features
- Can be outperformed by more complex models
Primary Use Cases
- Initial text classification
- Spam detection
- Sentiment analysis
Model Visualizations
Performance metrics and feature importance