Sensitivity analysis of parametric t-norm and s-norm based fuzzy classification system
Mingzhu Lu, Long Chen, C. L. Philip Chen · 2010
To solve the classification problems more adaptively and accurately, this paper studies the parametric t-norm and s-norm based fuzzy classification systems, where the fuzzy decision tree provides fuzzy rules and the system sensitivity on historical data works as a feedback controller. The system sensitivity with respect to the parameters of parametric t-norm and s-norm is investigated as the vehicle to reveal the regulation rules between the parameters and the system output. Based on the simulation results on several UCI data sets and the analysis of the system sensitivity, this paper provides some valuable regulation rules between the system output and the parameters of t-norm and s-norm for data with different characteristics. As the feedback controller, the system sensitivity, which is used to tune the parameters for unknown samples, improves the system's efficiency a lot.