Explaining dimensionality reduction results using Shapley values

• Clusters of dimensionality reduction results are interpreted using Shapley values. Summary visualizations convey the decisions processes. values can reveal ubiquitous information medical datasets. Dimensionality (DR) techniques have been consistently supporting high-dimensional data analysis in various applications. Besides patterns uncovered by these techniques, interpretation DR based on each feature’s contribution to low-dimensional representation supports new finds through exploratory analysis. Current literature approaches designed interpret do not explain features’ contributions well since they focus only or consider relationship among features. This paper presents ClusterShapley address problems, generate explanations and algorithms a cluster-oriented explains formation clusters meaning their relationship, which is useful for domains. We propose novel visualization guide clustering validate our methodology case studies publicly available The demonstrate approach’s interpretability power insights about pathologies patients different conditions results.