Decision-making process analysis and semantic explanation extraction from trained supervised machine learning models in medical expert systems

Abstract

Supervised machine learning provides a mechanism for establishing an approximation of input-output relationship between arbitary dataset. However, semantic interpretation of an underlying decision-making process of a trained model is very hard, especially considering the probabilistic nature of machine learning. The paper discusses possible ways to semantically explain decisionmaking process of a trained supervised machine learning model in order to gain insights to the dataset and derive new expert knowledge from such models.