.png){kind=icon}
.png){kind=icon}
Return
Augmenting Fuzzy Rule-Based Models through Random Sampling and Instance-Specific Model Learning
.png){kind=icon}
PRE
AI
DOI：10.1016/j.fss.2026.110002.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Fuzzy rule-based systems remain pivotal architectures of fuzzy models, yet persistent challenges in scalability and integration demand innovative solutions to unlock their full potential. In this study, we develop a framework that establishes a series of mapping relationships between local input and output data through randomly sampling the input space, including both data instances and feature. This strategy allows the framework to effectively capture diverse and localized input–output relationships, enhancing model flexibility and representation capacity. For each training sample, we evaluate the prediction performance across all models and assign the model with the smallest error as its dedicated model. This establishes a direct mapping between training samples and their most suitable predictive models. When predicting the output of a test sample, we employ a nearest-neighbor approach to identify the most similar training sample and use its dedicated model for prediction. This locally adaptive strategy leverages the strengths of sample-specific modeling while maintaining generalization ability. The effectiveness of the proposed scheme is validated through extensive experiments on a wide range of publicly available datasets.
Journal
IF:
2.7
Papers:
7.6K
Citations:
1.5W