Interaction Detection with Bayesian Decision Tree Ensembles

2018/09/23 by Junliang Du, Du, Junliang, Antonio R. Linero +1 · 2 citations
Computer Science · #Data Stream Mining Techniques #Neural Networks and Applications #Explainable Artificial Intelligence (XAI)

paper · pdf · doi:10.48550/arxiv.1809.08524

Abstract

Methods based on Bayesian decision tree ensembles have proven valuable in constructing high-quality predictions, and are particularly attractive in certain settings because they encourage low-order interaction effects. Despite adapting to the presence of low-order interactions for prediction purpose, we show that Bayesian decision tree ensembles are generally anti-conservative for the purpose of conducting interaction detection. We address this problem by introducing Dirichlet process forests (DP-Forests), which leverage the presence of low-order interactions by clustering the trees so that trees within the same cluster focus on detecting a specific interaction. We show on both simulated and benchmark data that DP-Forests perform well relative to existing interaction detection techniques for detecting low-order interactions, attaining very low false-positive and false-negative rates while maintaining the same performance for prediction using a comparable computational budget.

Cited by

• Purifying Interaction Effects with the Functional ANOVA: An Efficient Algorithm for Recovering Identifiable Additive Models
• Innovations in Multiple Sclerosis Care: The Impact of Artificial Intelligence via Machine Learning on Clinical Research and Decision-Making

Related

• Interpretable Selection and Visualization of Features and Interactions\n Using Bayesian Forests
• Pure interaction effects unseen by Random Forests
• DIRV: Dense Interaction Region Voting for End-to-End Human-Object Interaction Detection
• Fast and More Powerful Selective Inference for Sparse High-order Interaction Model
• Detecting Human-Object Interactions with Action Co-occurrence Priors