def overproductionProcess(folder): #metoda izdelana za testiranje metaDes, ne potrebujemo, ker pridobimo klasifikatorje od ostalih XProd, YProd = Helpers.readData() # divideDataForMeta(X, Y) #it divides data into Production, Meta and Selection XMeta, YMeta, XSel, YSel, XTest, YTest = readForMeta2(folder) overproductionELM(XProd,YProd, XMeta, XSel, XTest, folder=folder) #we generate classifiers and use them for responses overproductionRf(XProd,YProd, XMeta, XSel, XTest, folder=folder)
for alpha in alphas: for actFunction in activation_functions: cls = GenELMClassifier(hidden_layer = RandomLayer(n_hidden = n_hidden, activation_func = actFunction, alpha=alpha)) parameter = Helpers.cv(X,Y,cls,5, printing = False) parameter = parameter+ [n_hidden, alpha, actFunction, "normal"] parameters.append(parameter) print(parameter, "%d/%d" %(trial,nrOfTrials)) Helpers.pickleListAppend2(parameter, "parametersELM.p") # parameter = Helpers.cv(X,Y,BaggingClassifier(cls,n_estimators=30),10, printing = False) # parameter = parameter+ [n_hidden, alpha, actFunction, "bagged"] # parameters.append(parameter) # print(parameter, "%d/%d" %(trial,nrOfTrials)) trial = trial+1 # pickle.dump(parameters,open("parametersMultiQuadric.p","wb")) return def vrniMethod(self, parameter): # acc,prec,precTress, n_hidden, rhl, actFunction = parameter cls = GenELMClassifier(hidden_layer = RandomLayer(n_hidden = parameter[-3], activation_func = parameter[-1], alpha=parameter[-2])) return cls # tr, ts, trRaw, tsRaw, prec, precTress = res_dist(X,Y,cls,10) if __name__ == "__main__": X, Y = Helpers.readData() ELMMethod().poisciParametre(X,Y) # vrniMethod(pickle.load(open("parameter.p","rb")), X, Y)
def main1(): #method for testing MetaDES X, Y = Helpers.readData() divideDataForMeta(X,Y)