Identification of Fuzzy Rule-Based Models With Collaborative Fuzzy Clustering.

Area of Science:

• Computational Intelligence
• Machine Learning
• Data Mining

Background:

• Fuzzy rule-based models (FRBMs) are effective for complex systems.
• Data privacy concerns can prevent access to complete input-output datasets for FRBM development.
• Existing methods do not fully address FRBM identification under partial data ownership.

Purpose of the Study:

• To develop a strategy for building FRBMs when input or output data is inaccessible due to privacy constraints.
• To enable the identification of multiple-input-single-output (MISO) and multiple-input-multiple-output (MIMO) systems under privacy-preserving conditions.
• To leverage collaborative fuzzy clustering (CFC) for privacy-conscious FRBM construction.

Main Methods:

• Application of the collaborative fuzzy clustering (CFC) concept and algorithm.
• Utilizing structural information from input and output spaces via partition matrices for collaboration.
• Developing FRBMs for MISO and MIMO systems with decentralized data.

Main Results:

• FRBMs can be successfully built even when input and output data cannot be collected simultaneously.
• The collaborative mechanism facilitates structural information exchange, leading to more relevant system structures.
• Proposed approach demonstrates superior model performance compared to state-of-the-art methods on synthetic and real-world datasets.

Conclusions:

• The CFC-based approach effectively addresses FRBM identification challenges posed by data privacy.
• Collaborative FRBMs enhance model performance by enabling effective sharing of structural information between input and output spaces.
• This strategy offers a viable solution for developing FRBMs in privacy-sensitive, distributed data environments.