ACDC: $\alpha$-Carving Decision Chain for Risk Stratification

TL;DR

ACDC is a novel decision chain algorithm designed for risk stratification in healthcare, effectively handling large, imbalanced datasets and providing interpretable decision sequences with visual performance metrics.

Contribution

Introduces ACDC, a new decision chain method that sequentially isolates classes, improving interpretability and performance in imbalanced healthcare data analysis.

Findings

Effective in class-imbalanced health datasets

Provides interpretable decision chains with visual metrics

Outperforms traditional decision trees in risk stratification

Abstract

In many healthcare settings, intuitive decision rules for risk stratification can help effective hospital resource allocation. This paper introduces a novel variant of decision tree algorithms that produces a chain of decisions, not a general tree. Our algorithm, $\alpha$-Carving Decision Chain (ACDC), sequentially carves out "pure" subsets of the majority class examples. The resulting chain of decision rules yields a pure subset of the minority class examples. Our approach is particularly effective in exploring large and class-imbalanced health datasets. Moreover, ACDC provides an interactive interpretation in conjunction with visual performance metrics such as Receiver Operating Characteristics curve and Lift chart.