RCAR Framework: Building a Regularized Class Association Rules Model in a Categorical Data Space

Abstract

Regularized Class Association Rules (RCAR) is an algorithm which produces rules based classifier in a categorical data space. The main goal of RCAR algorithm is to build classifiers which are as accurate as the state of the art algorithms, while improving the interpretability and allowing end-users to maintain and understand its outcome easily and without statistical modeling background. In this work, first, we introduce the RCAR framework, second, we provide the main functions which extract Class Association Rules (CARs), prune irrelevant rules, and rank the conserved CARs according to a set of weights calculated for each CAR. The RCAR framework also consists of multiple visualization techniques that traces the steps of the model building according to its parameters, which facilitates the model elaboration and tuning parameter for simple users. Eventually, we implemented the RCAR algorithm in the RCAR's R package.