Data Complexity and Its Effect on EBRB System Accuracy

In recent years, the relationship between data quality and machine learning models has garnered increasing attention. The present work quantifies dataset quality using complexity measurements and conducts detailed experimental analysis on the impact of these measurements on the accuracy of the Extend Belief Rule Base (EBRB) model, a representative advanced rule-based system that has attracted great attention in the last few years. The purpose is to identify several complexity measurements that most significantly affect and are most sensitive to the accuracy of the EBRB system. In the experimental section, we validate the impact of complexity measurements on the EBRB model across 108 classification task datasets, and perform sensitivity analysis in determining the top five complexity measurements that most significantly impact the performance of the EBRB model. Accordingly, we provide guidance that can help decision-makers in the data preprocessing aspects of the EBRB model, thereby improving its performance and enhancing the accuracy of the EBRB model.