A fuzzy clustering-based rule model constructed with the regularization strategy and extreme learning machine

Nowadays, fuzzy rule models are widely used in solving classification problems. However, traditional fuzzy rule models often show certain limitations when faced with the characteristics of complex systems. In the present work, we put forth a fuzzy clustering-based rule model comprised of regularization strategies and the extreme learning machine, which can build proper local fuzzy rule models and achieve a balance between the computing overhead and the model’s interpretability without sacrificing the prediction performance. Specifically, the data structure is revealed by information granulation, and the input space is divided into local regions to form the premise module; Next, extreme learning machines are used in the conclusion module to enhance the description and representation capacities of the data behaviors within these local regions; Then, the data is classified by using the classification boundaries with the regularization strategy. Our proposed model is compared with reinforced fuzzy clustering-based rule models, and it is found that our model achieved a significantly higher accuracy in classification, which could partly be attributed to the randomness of the extreme machine in the proposed method. Moreover, introducing the regularization strategy to our model improves its generalization capacity and prediction performance, verifying its potential for broader applications.