Enhancing Graphical Interpretability for Functional Random Survival Forest

Functional Random Survival Forest (FRSF) has emerged as a powerful tool for analyzing functional data by starting from time-to- event data in survival analysis. However, interpreting and explaining this black-box model. This article deepens into methods for improving interpretability of Functional Survival Trees (FSTs) and explainability of FRSF . Introducing censored functional data (CFD), our approach aims to show the underlying mechanisms for survival predictions. We explore techniques for showing the significance of functional features and their contributions to survival outcomes, via the extension of the Permuta- tion Feature Importance (PFI) tool to functional principal component scores . Moreover, we extend the concept of separation curves to survival problems with suitable interpretations. This approach aims to elucidate the model’s decision-making process and highlight key factors influencing survival predictions.