def CrossValDE(modelName,
X,
y,
metric,
procsessor=None,
cv=3,
times=1,
random_state=0):
"""
优化DE时用的交叉验证
"""
res = []
yt = []
for i in y:
if (i == 0):
yt.append(0)
else:
yt.append(1)
for t in range(times):
skf = StratifiedKFold(n_splits=cv,
shuffle=True,
random_state=random_state + t)
indices = list(skf.split(X=X, y=yt))
# print (" ",time.strftime("%M:%S"))
for k in indices:
x_train, y_train, x_test, y_test = X[k[0]], y[k[0]], X[k[1]], y[
k[1]]
if (procsessor is not None):
# t=procsessor.transform(x_train,y_train)
# print("t",len(t))
# if(len(t)==3):
# qqq=1
# qqq+=1
x_train, y_train = procsessor.transform(x_train, y_train)
estimator = buildModel(x_train, y_train, modelName)
res.append(E.Eva(estimator, x_test, y_test, metric))
# print (" ",time.strftime("%M:%S"))
res = np.array(res)
return res.mean()
def fitness(self, targetPars, otherPars):
"""
对于SMOTUNED，otherPars有一个值，数据集
"""
res = []
data = deepcopy(otherPars[0])
stf = self.stratify(K=3, dataSet=data, randomSeed=1)
for i in range(3):
trainingSet = deepcopy(data.iloc[stf[i][0], :])
testSet = deepcopy(data.iloc[stf[i][1], :])
S = SMOTER(targetPars[0, 0], targetPars[0, 1], targetPars[0, 2],
trainingSet)
trainingSet_SMOTE = S.smoteR()
model = BM.buildModel(trainingSet=trainingSet_SMOTE,
modelType=self.modelName)
res.append(
E.Eva(model,
testSet=testSet,
metric=self.metric,
predValue="bug"))
return sum(res) / len(res)
# testSets=["ant-1.5","camel-1.2","ivy-2.0"]
trainingProjs=["ant","camel","jedit","synapse","ivy","xalan","xerces"]
unsupervisedModels=['U_LocRfc_A','U_LocRfc_D','U_amc_A','U_amc_D','U_avg_cc_A','U_avg_cc_D','U_ca_A','U_ca_D','U_cam_A','U_cam_D','U_cbm_A',
'U_cbm_D','U_cbo_A','U_cbo_D','U_ce_A','U_ce_D','U_dam_A','U_dam_D','U_dit_A','U_dit_D','U_ic_A','U_ic_D','U_lcom3_A',
'U_lcom3_D','U_lcom_A','U_lcom_D','U_loc_A','U_loc_D','U_max_cc_A','U_max_cc_D','U_mfa_A','U_mfa_D','U_moa_A','U_moa_D',
'U_noc_A','U_noc_D','U_npm_A','U_npm_D','U_rfc_A','U_rfc_D','U_wmc_A','U_wmc_D']
#i will give you a simple example about how to use our codes
# In[]Step 1 get traning set and test set
x_train,y_train,x_test,y_test=Data(paths=datasetsPaths, scenario="Cross-Version", testSetName="ant-1.7", trainingProjName="").getTrainingAndTestSet()
# In[] step 2 build and train a model
smo=SMOTEND(k=10, m=3, r=2)#you can use SMOTEND or SMOTEND+DE, NOTICE:the smotendde return the last generation
x_train,y_train=smo.transform(x_train,y_train)
m=BM.buildModel(x_train,y_train,modelName="RF")
# In[] step3 model selection
FPAResult=E.Eva(m,x_test,y_test,metric="FPA")
print(FPAResult)
def f2(trainingSet, testSet, modelNames, metrics, burak):
"""
不适用与SMOTERDE的测试
Only suitable to SMOTEND+DE
On given trainingSer and testSet, we return the metrcics' values for each models
@Parameters:
---------------------------
trainingSet:
train model
----------------------------------
testSet:
test set
---------------------------------
modelNames:
list-like, store each models' names
----------------------------------
metrics:
measures, lisst-like
-------------------------------------
burak:
whether use burak
-------------------------------------
smoter:
whether use smoter
--------------------------------------
rus:
whether use rus
-------------------------------------------
"""
metricValues, metricParas = [], []
for metric in metrics:
modelValues, modelParas = [], [] #存放每一个模型的值,存放每个模型的最佳参数
for model in modelNames:
maxValue, bestParas = -10.0, None
for ti in range(4):
dpp = DataPreProcessing(data=trainingSet,
targetData=testSet,
burakKNN=10,
burak=burak,
smoter=False,
rusRatio=75,
rus=False,
Drange=[
list(range(1, 21)),
[0, 1, 2, 3, 4, 5, 6], (0.1, 5)
],
metric=metric,
modelName=model,
F=0.7,
CR=0.3,
PopulationSize=10,
Lives=8,
smoterde=True)
newTrainingSet = dpp.preProcess()
mo = BM.buildModel(trainingSet=newTrainingSet, modelType=model)
t = E.Eva(model=mo,
testSet=testSet,
metric=metric,
predValue="bug")
if (maxValue < t):
maxValue = t
bestParas = dpp.bestParas
bestParas = str(bestParas[0][0]) + "," + str(
bestParas[0][1]) + "," + str(bestParas[0][2])
# maxValue=2
# bestParas="2,2,3"
#这里的maxValue就是针对特定 训练集、测试集、指标和模型四个维度的一个特定值
modelValues.append(maxValue)
modelParas.append(bestParas)
metricValues.append(modelValues)
metricParas.append(modelParas)
return (np.array(metricValues), np.array(metricParas))
def f1(trainingSet, testSet, modelNames, metrics, burak, smoter, rus):
"""
不适用与SMOTERDE的测试
Not suitable to SMOTEND+DE
On given trainingSer and testSet, we return the metrcics' values for each models
@Parameters:
---------------------------
trainingSet:
train model
----------------------------------
testSet:
test set
---------------------------------
modelNames:
list-like, store each models' names
----------------------------------
metrics:
measures, lisst-like
-------------------------------------
burak:
whether use burak
-------------------------------------
smoter:
whether use smoter
--------------------------------------
rus:
whether use rus
-------------------------------------------
"""
res = []
ms = []
dpp = DataPreProcessing(
data=trainingSet,
targetData=testSet,
burakKNN=10,
burak=burak,
smoter=smoter,
rusRatio=None,
rus=rus,
Drange=[list(range(5, 21)), [0, 1, 2, 3, 4, 5, 6], (0.1, 5)],
metric=None,
modelName=None,
F=0.7,
CR=0.3,
PopulationSize=10,
Lives=8,
smoterde=False)
newTrainingSet = dpp.preProcess()
for i in modelNames: #建立好所有模型
ms.append(BM.buildModel(trainingSet=newTrainingSet, modelType=i))
for metric in metrics:
t = []
for model in ms:
t.append(
E.Eva(model=model,
testSet=testSet,
metric=metric,
predValue="bug"))
res.append(t)
return np.array(res)
