Improve prediction accuracy of your fuzzy systems: A simulated annealing approach

Abstract

For various physical processes, especially those demanding high cost or operational time, it becomes crucial to have accurate predictions of their key performance measures based on given settings of different input parameters. Among other artificial intelligence based tools, fuzzy rule-based systems have also been widely used for this purpose. Widespread applicability of the fuzzy systems has been restricted by the lack of accuracy in the prediction results and inherent difficulties in different approaches that have been utilized for improving their prediction capabilities. The prediction accuracy of a fuzzy rule-based system is often lost because of non-optimal design of fuzzy sets and/or incorrect combination of these fuzzy sets in development of fuzzy rules. The exhaustive search for finding the optimal design of fuzzy sets or optimal formation of rule-base is practically unfeasible. In order to address this issue, the authors have put forward a search algorithm (along with its application) that provides optimal design of fuzzy sets as well as optimal combination of fuzzy sets in a rule-base. It has been detailed in this chapter that the application of this algorithm significantly increases the prediction accuracy of a fuzzy rule-based system. The presented work possesses immense potential to be utilized at practical levels, in various fields, for effective modeling of physical phenomena and accurate predictions of the performance measures. In the current work, a two phase methodology for optimal formation of fuzzy rule-base is presented. A rule-base is developed from experimental data related to a physical process. The developed rule-base is expected to predict the values of the response variable(s) of the process based on the values of the predictor variables. In the first phase, the formation of the fuzzy rules is optimized by assigning the most appropriate fuzzy sets of the response variable to the rules. The task is accomplished by utilizing a combinatorial optimization method called as Simulated Annealing Algorithm. In the second phase, the effectiveness of the rule-base is further improved by adjusting the fuzzy sets of the response variable and again Simulated Annealing Algorithm has been utilized for this purpose. The fuzzy rule-base is developed from the data related to turning process utilized for machining of tool steel. The same physical process (turning) has been used for validation tests of the optimized fuzzy rule-base. © 2012 Nova Science Publishers, Inc. All rights reserved.