Machine Learning and Inference Reporting Criteria

(Wojtusiak)

The goal of this project is to design reporting criteria for Machine Learning applications in areas in which special care need to be taken in terms of transparency and reproducibility of results. The following ten criteria have been developed:

MLI-1: Experimental Design

MLI-2: Statistical Model Evaluation

MLI-3: Model Calibration

MLI-4: Top Predictors

MLI-5: Global Sensitivity Analysis

MLI-6: Decision Curve Analysis

MLI-7: Global Model Explanation

MLI-8: Local Prediction Explanation

MLI-9: Programming Interface

MLI-10: Source Code