Optimization of the ANFIS using a genetic algorithm for physical work rate classification

Abstract

Purpose. Recently, a new method was proposed for physical work rate classification based on an adaptive neuro-fuzzy inference system (ANFIS). This study aims to present a genetic algorithm (GA)-optimized ANFIS model for a highly accurate classification of physical work rate. Methods. Thirty healthy men participated in this study. Directly measured heart rate and oxygen consumption of the participants in the laboratory were used for training the ANFIS classifier model in MATLAB version 8.0.0 using a hybrid algorithm. A similar process was done using the GA as an optimization technique. Results. The accuracy, sensitivity and specificity of the ANFIS classifier model were increased successfully. The mean accuracy of the model was increased from 92.95 to 97.92%. Also, the calculated root mean square error of the model was reduced from 5.4186 to 3.1882. The maximum estimation error of the optimized ANFIS during the network testing process was ± 5%. Conclusion. The GA can be effectively used for ANFIS optimization and leads to an accurate classification of physical work rate. In addition to high accuracy, simple implementation and inter-individual variability consideration are two other advantages of the presented model.

Keywords:

• physical work rate
• classification
• optimization
• adaptive neuro-fuzzy inference system

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by Isfahan University of Medical Sciences [grant number 394796].