anfis.zip_ANFIS_anfis code_anfis fuzzy_fuzzy ANFIS_zip

load Dataset/glass.data; load Dataset/wine.data; addpath Functions/; nglass = NormalizeDataset(glass); nwine = NormalizeDataset(wine); save SavedData/NormalizedGlass.data nglass -ascii; save SavedData/NormalizedWine.data nwine -ascii; RulebaseGlassInitial = GenerateFuzzyRulesMaxL2(nglass); RulebaseWineInitial = GenerateFuzzyRulesMaxL2(nwine); save SavedData/RulebaseGlassInitial.data RulebaseGlassInitial -ascii; save SavedData/RulebaseWineInitial.data RulebaseWineInitial -ascii; %======= load SavedData/NormalizedGlass.data; load SavedData/NormalizedWine.data; load SavedData/RulebaseGlassInitial.data; load SavedData/RulebaseWineInitial.data; GlassData = NormalizedGlass(:,1:9); GlassClass = NormalizedGlass(:,10); WineData = NormalizedWine(:,1:13); WineClass = NormalizedWine(:,14); ReducedGlass40 = ChooseQForEachClass(RulebaseGlassInitial,... GlassData,GlassClass,40); ReducedGlass100 = ChooseQForEachClass(RulebaseGlassInitial,... GlassData,GlassClass,100); ReducedWine100 = ChooseQForEachClass(RulebaseWineInitial,... WineData,WineClass,100); save SavedData/ReducedGlass40.data ReducedGlass40 -ascii; save SavedData/ReducedGlass100.data ReducedGlass100 -ascii; save SavedData/ReducedWine100.data ReducedWine100 -ascii; %======== % load SavedData/NormalizedGlass.data; load SavedData/ReducedGlass40.data addpath Functions/; dglass = NormalizedGlass(:,1:9); cglass = NormalizedGlass(:,10); for i=1:3 [finalRulebase,~,error_rate] = OptimizationSingleWinner(ReducedGlass40,dglass,cglass); x = linspace(i-1,i,size(ReducedGlass40,1)); error_rate = error_rate'; plot(x,error_rate); hold on; ReducedGlass40 = finalRulebase; end xlabel('iteration number'); ylabel('error rate'); title('Error-rate on training data with single winner reasoning method'); print -djpeg Figures/Error_rateSW.jpg; save SavedData/FinalRuleBaseGlass40_3iterationsSW finalRulebase -ascii; %====================== load SavedData/NormalizedGlass.data; load SavedData/NormalizedWine.data; load SavedData/ReducedGlass100.data; load SavedData/ReducedWine100.data; addpath Functions/; dglass = NormalizedGlass(:,1:9); cglass = NormalizedGlass(:,10); dwine = NormalizedWine(:,1:13); cwine = NormalizedWine(:,14); [finalRulebaseGlassSW,accGlass,~] = OptimizationSingleWinner(ReducedGlass100,dglass,cglass); [finalRulebaseWineSW,accWine,~] = OptimizationSingleWinner(ReducedWine100,dwine,cwine); GlassAccNumberOfRulesSW = [accGlass(size(ReducedGlass100,1))*100 size(finalRulebaseGlassSW,1)]; WineAccNumberOfRulesSW = [accWine(size(ReducedWine100,1))*100 size(finalRulebaseWineSW,1)]; save SavedData/WineAccNumberOfRulesSW.data WineAccNumberOfRulesSW -ascii; save SavedData/GlassAccNumberOfRulesSW.data GlassAccNumberOfRulesSW -ascii; save SavedData/finalRulebaseGlassSW.data finalRulebaseGlassSW -ascii; save SavedData/finalRulebaseWineSW.data finalRulebaseWineSW -ascii; %===================== load SavedData/NormalizedGlass.data; load SavedData/ReducedGlass40.data addpath Functions/; dglass = NormalizedGlass(:,1:9); cglass = NormalizedGlass(:,10); for i=1:3 [finalRulebase,~,error_rate] = OptimizationWeightedVote(ReducedGlass40,dglass,cglass); x = linspace(i-1,i,size(ReducedGlass40,1)); error_rate = error_rate'; plot(x,error_rate); hold on; ReducedGlass40 = finalRulebase; end xlabel('iteration number'); ylabel('error rate'); title('Error-rate on training data with weighted vote reasoning method'); print -djpeg Figures/Error_rateVW.jpg; save SavedData/FinalRuleBaseGlass40_3iterationsWV finalRulebase -ascii; %====================== load SavedData/NormalizedGlass.data; load SavedData/NormalizedWine.data; load SavedData/ReducedGlass100.data; load SavedData/ReducedWine100.data; addpath Functions/; dglass = NormalizedGlass(:,1:9); cglass = NormalizedGlass(:,10); dwine = NormalizedWine(:,1:13); cwine = NormalizedWine(:,14); [finalRulebaseGlassWV,accGlass,~] = OptimizationWeightedVote(ReducedGlass100,dglass,cglass); [finalRulebaseWineWV,accWine,~] = OptimizationWeightedVote(ReducedWine100,dwine,cwine); GlassAccNumberOfRulesWV = [accGlass(size(ReducedGlass100,1))*100 size(finalRulebaseGlassWV,1)]; WineAccNumberOfRulesWV = [accWine(size(ReducedWine100,1))*100 size(finalRulebaseWineWV,1)]; save SavedData/WineAccNumberOfRulesWV.data WineAccNumberOfRulesWV -ascii; save SavedData/GlassAccNumberOfRulesWV.data GlassAccNumberOfRulesWV -ascii; save SavedData/finalRulebaseGlassWV.data finalRulebaseGlassWV -ascii; save SavedData/finalRulebaseWineWV.data finalRulebaseWineWV -ascii; %========================= load SavedData/NormalizedGlass.data; load SavedData/NormalizedWine.data; addpath Functions/; GlassData = NormalizedGlass(:,1:9); GlassClass = NormalizedGlass(:,10); WineData = NormalizedWine(:,1:13); WineClass = NormalizedWine(:,14); MGlassData = size(GlassData,1); MWineData = size(WineData,1); finalRulebaseLengthSW = 0; finalRulebaseLengthWV = 0; accTestSW = 0; accTestWV = 0; for s = 1:MGlassData dglass = [GlassData(1:s-1,:) ; GlassData(s+1:MGlassData,:)]; cglass = [GlassClass(1:s-1,:) ; GlassClass(s+1:MGlassData,:)]; RulebaseGlassInitial = GenerateFuzzyRulesMaxL2([dglass cglass]); ReducedGlass100 = ChooseQForEachClass(RulebaseGlassInitial,... dglass,cglass,100); [finalRulebaseGlassSW,~,~] = ... OptimizationSingleWinner(ReducedGlass100,dglass,cglass); sclass = ClassifyTestSingleWinner(dglass(s,:),finalRulebaseGlassSW,6); if sclass == cglass(s) accTestSW = accTestSW + 1; end finalRulebaseLengthSW = finalRulebaseLengthSW + size(finalRulebaseGlassSW,1); [finalRulebaseGlassVW,~,~] = ... OptimizationWeightedVote(ReducedGlass100,dglass,cglass); sclass = ClassifyTestWeightedVote(dglass(s,:),finalRulebaseGlassVW,6); if sclass == cglass(s) accTestWV = accTestWV + 1; end finalRulebaseLengthWV = finalRulebaseLengthWV + size(finalRulebaseGlassVW,1); end accTestSW = (accTestSW / MGlassData) * 100; finalRulebaseLengthSW = finalRulebaseLengthSW/MGlassData; accTestWV = (accTestWV / MGlassData) * 100; finalRulebaseLengthWV = finalRulebaseLengthWV/MWineData; save SavedData/accTestGlassSW.data accTestSW -ascii; save SavedData/numberOfRulesTestGlassSW.data finalRulebaseLengthSW -ascii; save SavedData/accTestGlassWV.data accTestWV -ascii; save SavedData/numberOfRulesTestGlassWV.data finalRulebaseLengthWV -ascii; finalRulebaseLengthSW = 0; finalRulebaseLengthWV = 0; accTestSW = 0; accTestWV = 0; for s = 1:MWineData dwine = [WineData(1:s-1,:) ; WineData(s+1:MWineData,:)]; cwine = [WineClass(1:s-1,:) ; WineClass(s+1:MWineData,:)]; RulebaseWineInitial = GenerateFuzzyRulesMaxL2([dwine cwine]); ReducedWine100 = ChooseQForEachClass(RulebaseWineInitial,... dwine,cwine,100); [finalRulebaseWineSW,~,~] = ... OptimizationSingleWinner(ReducedWine100,dwine,cwine); sclass = ClassifyTestSingleWinner(dwine(s,:),finalRulebaseWineSW,3); if sclass == cwine(s) accTestSW = accTestSW + 1; end finalRulebaseLengthSW = finalRulebaseLengthSW + size(finalRulebaseWineSW,1); [finalRulebaseWineWV,~,~] = ... OptimizationWeightedVote(ReducedWine100,dwine,cwine); sclass = ClassifyTestWeightedVote(dwine(s,:),finalRulebaseWineWV,3); if sclass == cwine(s) accTestWV = accTestWV + 1; end finalRulebaseLengthWV = finalRulebaseLengthWV + size(finalRulebaseWineWV,1); end accTestSW = (accTestSW / MGlassData) * 100; finalRulebaseLengthSW = finalRulebaseLengthSW/MWineData; accTestWV = (accTestWV / MGlassData) * 100; finalRulebaseLengthWV = finalRulebaseLengthWV/MWineData; save SavedData/accTestWineSW.data accTestSW -ascii; save SavedData/numberOfRulesTestWineSW.data finalRulebaseLengthSW -ascii; save