Related Experiment Videos
Refinement of generated fuzzy production rules by using a fuzzy neural network
Eric C C Tsang1, Daniel S Yeung, John W T Lee
1Department of Computing, The Hong Kong Polytechnic University, Kowloon, Hong Kong. csetsang@comp.polyu.edu.hk
Summary
This study enhances fuzzy production rules (FPRs) using a fuzzy neural network (FNN) to improve classification accuracy for unseen data. The method refines rule weights, reducing expert consultation time and enhancing knowledge representation.
Area of Science:
- Artificial Intelligence
- Machine Learning
- Fuzzy Systems
Background:
- Fuzzy production rules (FPRs) are used to represent uncertain domain knowledge in fuzzy systems.
- Current methods for obtaining FPRs, like expert interviews or machine learning, yield suboptimal, redundant rules with low accuracy on unseen data.
Purpose of the Study:
- To address limitations of existing FPRs by enhancing their representation power.
- To develop a fuzzy neural network (FNN) with an improved learning algorithm for FPR refinement.
- To improve the accuracy of classifying unseen samples using refined FPRs.
Main Methods:
- Enhanced FPRs by incorporating local and global weights.
- Developed a fuzzy neural network (FNN) with an advanced learning algorithm.
- Utilized the FNN to refine the local and global weights of FPRs.
Main Results:
- The proposed method achieved high accuracy in classifying unseen samples.
- The number of extracted FPRs did not increase.
- The time required for domain expert consultation was significantly reduced.
Conclusions:
- The enhanced FPRs and FNN-based refinement method effectively improve classification accuracy.
- The approach offers a more efficient way to extract and refine domain knowledge compared to traditional methods.