No Detail Found!

Effectiveness of enhanced takagi sugeno kang's fuzzy inference model

Author(s)
SUGIYARTO SURONO Department of Mathematics, Ahmad Dahlan University,Yogyakarta,Indonesia
GOH KHANG WEN Faculty of Data Science and Information Technology, INTI International University,Nilai,Malaysia
CHOO WOU ONN Faculty of Data Science and Information Technology, INTI International University,Nilai,Malaysia
YOSZA BIN DASRIL Faculty of Technology Management and Business, Universiti Tun Husseion Onn Malaysia, Batu Pahat, Johor, Malaysia
ASIH YULI ASTUTI Department of Mathematics, Ahmad Dahlan University,Yogyakarta,Indonesia
NITHYA PERIASAMY Department of Mathematics,Science and Computer, Ungku Omar Polytechnic, Perak,Malaysia
Abstract
Fuzzy logic is one of the components from soft computing and it is useful as a method to map the problem from input into the desired output. The main feature of fuzzy logic is membership function. Takagi Sugeno Kang fuzzy system appears to be more preferred and easier to use, thanks to its simple structure and high proximity. The use of learning rate is extremely important in Mini-Batch Gradient Descent to improve the quality and the rate of training convergence. However, choosing the right learning rate is one of the shortcomings of the MBGD method, so AdaBound is used to optimize the selection of learning rates. AdaBound is an adaptive method that applies dynamic boundaries on learning rate. The said dynamic boundaries has the definition of learning rate limitation, both from up and down, to make sure the learning rate is neither too big nor too small. Aside from that, the boundaries become tighter as iteration goes and forcing the learning rate close to constant value. The optimization of AdaBound is intended not only to improve the convergence rate, but also to optimize the convergence into global minimum value in the end of training period. The method used in this study is Takagi Sugeno Kang (TSK) method. The rule of Takagi Sugeno Kang fuzzy system will be optimized using Mini-Batch Gradient Descent that has been modified by AdaBound. The goal of this study is to observe the accuracy of classification done by TSK with MBGD-A. The data used in this study is obesity data obtained from Kaggel dataset, while the variables are composed of three independent variables and one dependent variable. There are 32 rules in the TSK inference model based on data, however only 24 of them can be applied. To obtain firm numbers, the defuzzification value is derived using the weighted average approach from the 24 rules obtained. MAD was utilized as the evaluation model, and the error value was 14,549, indicating that the approach was effective and efficient.
Publication Details
Page(s) 2383-2392
DOI DOI not available
Published Journal: Journal of Theoretical and Applied Information Technology, Volume: 100, Issue: 8, Year: 2022

Request the full text from PASTIC

Keywords
optimization, MBGD, AdaBound, Fuzzy TSK
0

Citations

0

Downloads

55

Views