Agentic data science with the Data Agent Kit
cloud.google.com·6h ago
AI Research
arXiv CS.AI·Aug 15·5 min read
Personalized Scorer Modeling: A Learning-Based Framework for Deriving Robust Sleep Stage Labels from Multiple Experts
TL;DR
A learning-based framework was developed to improve sleep stage classification by aggregating labels from multiple experts. This approach, called personalized scorer modeling, achieved up to 86.07% accuracy, enhancing the reliability of sleep stage labels.
✦ Why It Matters
Implement personalized scorer modeling to improve sleep stage classification accuracy in your projects today.
Key Takeaways
- Utilize multi-scored datasets for better reference labels.
- Apply confusion matrices to model expert behavior.
- Experiment with random forest classifiers for optimal results.
- Aggregate expert labels to enhance classification reliability.
How It Works
The personalized scorer modeling framework uses confusion matrices to estimate the probability of true sleep stages based on multiple experts' labels. By normalizing these matrices and aggregating the probabilities, the framework assigns a final label for each sleep epoch, effectively capturing the collective judgment of the experts.