Dr. Nobias Fünke’s 100% Natural Good-Time Family Bias Solution
Metric Lattice for Performance Estimation (MLPE)
Metric Lattice for Performance Estimation (MLPE) directly addresses clinicians’ need to know when they can trust healthcare-related machine learning (ML) recommendations. MLPE takes a binary ML model for which classification thresholds have already been determined and provides clinicians with patient-specific model performance estimates. The tool produces per-patient confidence intervals of sensitivity and specificity at prediction time, which clinicians can use at the bedside to determine how much to trust a particular model output. If the confidence interval on model performance is too wide or low, the clinician may be advised against using the ML model for the patient in question, as the model prediction and corresponding recommendation may be biased or misleading.
Below is a video demonstration of our tool.
Below is a link to our GitHub Repository. Besides the code a readme.md and requirements.txt file are included.