def poisciParametre(self, X,Y):
parameterFile = "BaggingParametersCV.p"
# Helpers.deleteFile(parameterFile)
parameters = []
n_folds = 4
for n_est in [1,20,40,60,80,100,120]:
k_fold = cross_validation.KFold(n=X.shape[0], n_folds=n_folds, shuffle=True)
ratioCuts = [0,0.4,0.6,0.8,0.9,0.95, 1-2.0/n_est]
resu = np.zeros(len(ratioCuts))
for trainIdx, testIdx in k_fold:
baggedEst = Bagging(self.estimator, n_estimators=n_est)
baggedEst.fit(X[trainIdx],Y[trainIdx])
i=0
for ratioCut in ratioCuts: #n*(1-x) = 2, we select best 2 algos
# printSome(baggedEst,X[testIdx])
baggedEst.ratioCutOffEstimators = ratioCut
preds = baggedEst.predict_proba(X[testIdx])
res = Helpers.metricWithRawDataAboveDecil(preds[:,1],Y[testIdx],0.9, precision_score)
resu[i]+=res
i+=1
resu /= n_folds
for i, r in enumerate(resu):
parameter = (r,n_est,ratioCuts[i])
print(np.array(parameter)) #we put in array just for pretty print
parameters.append(parameter)
