Researchers have introduced the twoblock clustering tree (tbtree), a novel regression tree designed for multivariate responses. This method utilizes dense or sparse twoblock dimension reduction for local leaf models and impurity calculations, ensuring computational efficiency and interpretability. The paper also presents a coskewness-based estimator for the twoblock dimension reduced space, which aids in identifying non-normal clusters. While adept at recovering piecewise linear regimes, tbtree has demonstrated its capability to model complex nonlinear dependencies, performing comparably to techniques like random forests on real-world data. AI
IMPACT Introduces a new interpretable method for regression trees that may offer advantages over existing black-box models in specific applications.
RANK_REASON The cluster contains an academic paper detailing a new statistical methodology. [lever_c_demoted from research: ic=1 ai=0.4]
- arXiv
- Coskewness
- decision tree
- multivariate linear models
- random forest
- stat.ML
- Twoblock clustering tree
- twoblock dimension reduction
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
