A Comparative Study on Fuzzy Mamdani-Sugeno-Tsukamoto for the Childhood Tuberculosis Diagnosis

摘要

World Health Organization (WHO) estimated that approximately 80 thousand children died every year in view of Childhood Tuberculosis. The disease needs an appropriate treatment considering the difficulties in establishing a diagnosis in pediatric patients. The incapability of children to produce sputum becomes one of the difficulties. Sputum is used to diagnose a person suffering from tuberculosis, based on Mycobacterium tuberculosis in sputum. In this paper, Mamdani, Tsukamoto and Sugeno-types Fuzzy Inference System are applied to assist the tuberculosis diagnosis. The different technique in these three methods is aimed to determine the most appropriate method for such diagnosis. The results show that, of the three types of Fuzzy Inference System, the best model is Sugeno model. Sugeno-type FIS has a better accuracy compared to both Mamdani and Tsukamoto ones at 93%, equivalent to a fault diagnosis in 13 of 180 patients. Here, Mamdani-type FIS is provided the diagnostic accuracy of 89%, equivalent to the fault diagnosis in 20 of 180 patients. On the other hand, Tsukamoto is provided the diagnostic accuracy of 92%, equivalent to fault diagnosis in 15 of 180 patients. Based on the three systems, the most precise output is found in Sugeno-type Fuzzy with a value by 95.1% while for Fuzzy Mamdani and Tsukamoto, it values are 93.4% and 94.5%, respectively. Also, the highest level for the system sensitivity is found in Sugeno with 97.2% in comparison to Tsukamoto FIS by 96.67% and Mamdani at 94.4%.