Design of a TSK Rule-Based Model With Granular Rules and Ensemble Learning in Big Data

Nowadays, the management and analysis of big data have become major challenges for researchers in the field of data mining. The increasing rate of data generation, along with the need to extract meaningful patterns, highlights the necessity of developing scalable big data analysis methods. In this context, fuzzy rule-based models have emerged as powerful tools for knowledge extraction from data. However, designing these models typically requires the monolithic use of the entire dataset, which is impractical for big data scenarios due to computational limitations. This study introduces a novel concept and proposes a new framework for designing and evaluating the performance of rule-based models in big data environments. Within this framework, a set of rule-based submodels are randomly constructed using sampled data and trained through Bagging. The rules extracted from these submodels are then aggregated using an optimization-based weighting strategy combined with an information entropy method. This approach, which has not yet been explored in the literature, contributes to improving model efficiency. In the experimental section, large-scale datasets with high dimensionality and volume are employed to comprehensively evaluate the performance of the proposed model. The results demonstrate that the proposed model achieves significant improvements over comparable models.