Grafting for Combinatorial Boolean Model using Frequent Itemset Mining

TL;DR

This paper presents GRAB, an efficient algorithm for learning the combinatorial Boolean model from data by reducing the problem to frequent itemset mining, improving computational efficiency and interpretability.

Contribution

The paper introduces GRAB, a novel algorithm that enables scalable learning of CBM through reduction to frequent itemset mining, addressing previous computational challenges.

Findings

GRAB outperforms existing methods in computational efficiency.

GRAB achieves high prediction accuracy on benchmark datasets.

GRAB effectively discovers interpretable Boolean patterns.

Abstract

This paper introduces the combinatorial Boolean model (CBM), which is defined as the class of linear combinations of conjunctions of Boolean attributes. This paper addresses the issue of learning CBM from labeled data. CBM is of high knowledge interpretability but na\"{i}ve learning of it requires exponentially large computation time with respect to data dimension and sample size. To overcome this computational difficulty, we propose an algorithm GRAB (GRAfting for Boolean datasets), which efficiently learns CBM within the $L_1$-regularized loss minimization framework. The key idea of GRAB is to reduce the loss minimization problem to the weighted frequent itemset mining, in which frequent patterns are efficiently computable. We employ benchmark datasets to empirically demonstrate that GRAB is effective in terms of computational efficiency, prediction accuracy and knowledge discovery.