Post-hoc Interpretation of Clinical Pathways Clustering using Bayesian Inference
Ksenia Balabaeva, Sergey Valerevich Kovalchuk · Procedia Computer Science · 2020
This study is dedicated to the domain of explainable artificial intelligence. We propose an approach to machine learning clustering results interpretation and apply it to clinical clusters interpretation. We use Bayesian inference to the post-hoc interpretation of clustering provided by K-means algorithm. We investigate what are the differences and similarities between clusters comparing posterior distributions of features. The proposed approach is not model-specific and may be used for interpretation of any clustering algorithm. Finally, we compare the results with a medical expert interpretation and analyze the differences and similarities.