treeheatr

treeheatr integrates decision tree structures with heatmaps to visualize and interpret the distribution and correlation of features within terminal (leaf) nodes of tree models.

Key Features:

• Heatmap-Integrated Visualization: Combines decision tree structures with heatmap representations at terminal (leaf) nodes to display data distributions within each node.
• Interpretability and Correlation Analysis: Facilitates examination of feature correlation structure and highlights individual feature contributions to tree predictions.
• Customizable Visualizations and Input Types: Leverages the ggparty package for drawing decision trees and accepts input as a dataframe or tibble, or as precomputed tree objects (party or constparty) and customized tree nodes (partynode).

Scientific Applications:

• Model Interpretation: Visualizes how decision trees partition feature space and supports visual assessment of model behavior and performance.
• Feature Importance Analysis: Aids identification of influential features by displaying their distributions and correlations within terminal nodes.
• Educational Tool: Provides combined tree-and-heatmap visualizations to aid understanding of simple decision tree models.

Methodology:

Integration of decision tree structures with heatmaps at terminal nodes is implemented using the ggparty package.