Table 1 Classification accuracy of the automated approach
From: Epistemonikos: a comprehensive database of systematic reviews for health decision-making
Measure | Value (95% CI) |
|---|---|
Machine learning classifier (for records with abstract) | |
Sensitivity | 96.8% (96.58 to 97.06%) |
Specificity | 80.4% (79.55 to 81.23%) |
Precision | 92.2% (91.79 to 92.51%) |
Accuracy | 92.0% (91.65 to 92.28%) |
Heuristic classifier (for records without abstract) | |
Sensitivity | 94.1% (88.74 to 96.97%) |
Specificity | 55.6% (50.49 to 60.63%) |
Precision | 43.9% (38.34 to 49.71%) |
Accuracy | 66.0% (61.74 to 70.02%) |