Identification of representative trees in random forests based on a new tree-based distance measure

Abstract

In life sciences, random forests are often used to train predictive models. However, gaining any explanatory insight into the mechanics leading to a specific outcome is rather complex, which impedes the implementation of random forests into clinical practice. By simplifying a complex ensemble of decision trees to a single most representative tree, it is assumed to be possible to observe common tree structures, the importance of specific features and variable interactions. Thus, representative trees could also help to understand interactions between genetic variants. Intuitively, representative trees are those with the minimal distance to all other trees, which requires a proper definition of the distance between two trees. Thus, we developed a new tree-based distance measure, which incorporates more of the underlying tree structure than other metrics. We compared our new method with the existing metrics in an extensive simulation study and applied it to predict the age at onset based on a set of genetic risk factors in a clinical data set. In our simulation study we were able to show the advantages of our weighted splitting variable approach. Our real data application revealed that representative trees are not only able to replicate the results from a recent genome-wide association study, but also can give additional explanations of the genetic mechanisms. Finally, we implemented all compared distance measures in R and made them publicly available in the R package timbR (https://github.com/imbs-hl/timbR).

Explainable artificial intelligence in emergency medicine: an overview

“…Simplification creates a straightforward, less complicated interpretable model from a black-box model [ 22 ]. One example of simplification is selection of a single decision tree as the representative of a random forest ensemble of numerous decision tree models [ 40 , 41 ]. A simplified model could aggregate predictions from the individual trees to produce a final output, as shown in Fig.…”

“…Although this approach is commonly utilized in medical research, the results can be challenging to interpret due to the ensemble nature. Identifying a single tree that captures the primary patterns and behaviors of the entire forest allows a balance between interpretability and performance [ 40 , 41 ]. This representative tree can be visualized, providing insights into the decision-making process using the same foundational logic as the original ensemble.…”

“…(C) The bar plot displays the absolute SHAP values, indicating that factors of time to hospital, witnessed status, and age are more relevant predictors of survival than age across all cases. ance between interpretability and performance [40,41]. This representative tree can be visualized, providing insights into the decision-making process using the same foundational logic as the original ensemble.…”