Fuzzy rule based classification with FeatureSelector and modified threshold accepting

摘要: Abstract This paper highlights the need to reduce dimension of feature space in classification problems high dimensions without sacrificing power considerably. We propose a methodology for tasks which comprises three phases: (i) selection, (ii) automatic generation fuzzy if–then rules and (iii) reduction rule base while retaining its power. The first phase is executed by using FeatureSelector, software developed solely extraction pattern recognition problems. time that FeatureSelector used as preprocessor based systems. In second phase, standard system modified invoked with most important features extracted new set features. third threshold accepting algorithm (MTA), proposed elsewhere authors (Ravi et al., Fuzzy Sets Systems, forthcoming) minimizing number guaranteeing are taken objectives this multi objective combinatorial global optimization problem. here has been successfully demonstrated two well-known wine problem, includes 13 variables original form Wisconsin breast cancer determination 9 variables. conclusion, results encouraging there no remarkable both problems, despite fact some have deleted from study resorting selection. Also, MTA outperformed test considered here. suggest having higher can be solved within framework presented powers obtained when working less than significant achievement study.