ROPAC: Rule OPtimized Aggregation Classifier

Abstract

In the era of data-driven decision-making, extracting meaningful insights from vast amounts of information is paramount. In organizing this data, classification methods play a pivotal role. Among the existing classification techniques, rule-based classifiers have gained prominence for their effectiveness and interpretability. One such is the Rule Aggregation ClassifiER (RACER), known for its exceptional performance but limited when dealing with high-dimensional, low-sample-size datasets. In this paper, we introduce the Rule OPtimized Aggregation Classifier (ROPAC) as an extension of RACER that incorporates two different rule optimization methods, resulting in ROPAC-L and ROPAC-M, which aim to improve overall performance and classification accuracy. We evaluated this algorithm through experimentation with fifty datasets from reputable sources, such as the OpenML website and the UCI Machine Learning Repository. Furthermore, the proposed algorithm’s accuracy is compared with fifteen well-known classifiers. Our results demonstrate that ROPAC outperforms all the other algorithms in terms of accuracy, showcasing its superiority and dominance in various data scenarios.