ANFIS Construction With Sparse Data via Group Rule Interpolation.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Computational Intelligence

Background:

• Adaptive-network-based fuzzy inference systems (ANFIS) typically require substantial training data.
• Real-world applications often present challenges with limited available data.
• Existing ANFIS models may not perform optimally under data scarcity.

Purpose of the Study:

• To develop an automated construction method for ANFIS.
• To enhance Takagi-Sugeno fuzzy regression models for limited data scenarios.
• To improve the applicability of ANFIS in data-scarce environments.

Main Methods:

• Interpolation of fuzzy rules across domains using existing ANFIS.
• Creation of a rule dictionary by extracting rules from established ANFIS.
• Utilizing local linear embedding for rule interpolation and intermediate ANFIS construction.
• Employing a fine-tuning mechanism to refine the constructed ANFIS.

Main Results:

• The proposed approach significantly improves ANFIS performance in data shortage situations.
• Experimental evaluations on synthetic and real-world datasets validate the method's effectiveness.
• Demonstrated ability to overcome limitations of traditional ANFIS modeling with insufficient data.

Conclusions:

• The automated ANFIS construction method effectively addresses data scarcity challenges.
• This approach enhances the robustness and applicability of fuzzy inference systems.
• The technique offers a viable solution for ANFIS modeling in data-limited domains.