Automatic and heuristic complete design for ANFIS classifier

Area of Science:

• Artificial Intelligence
• Machine Learning
• Data Mining

Background:

• Adaptive Neuro-Fuzzy Inference System (ANFIS) is a popular fuzzy classifier.
• Designing ANFIS classifiers faces challenges in selecting membership functions and training methods to minimize classification errors.

Purpose of the Study:

• To present a novel intelligent method for optimizing ANFIS classifiers.
• To simultaneously select and locate membership functions and train the ANFIS model.
• To minimize classification errors in ANFIS using a new optimization approach.

Main Methods:

• A new technique based on intelligent methods is presented.
• Inclined Planes System Optimization (IPO) is utilized for simultaneous selection and location of membership functions and ANFIS training.
• The method is evaluated on diverse datasets with varying classes and feature vector lengths.

Main Results:

• The proposed method demonstrates higher accuracy and efficiency in selecting membership functions and simultaneous training compared to existing algorithms.
• The IPO-based method outperforms Particle Swarm Optimization, Genetic Algorithm, Differential Evolution, and ACOR algorithms.
• Effective classification of complex datasets with different characteristics is achieved.

Conclusions:

• The novel IPO-based intelligent method offers a superior approach for ANFIS classifier design.
• This technique effectively addresses the challenges of membership function selection and simultaneous training.
• The findings suggest significant improvements in ANFIS performance and classification accuracy.