def Train_Model(self):
print("Now, It is training all of this set!")
util_feature = FeatureExtraction.Util_Feature_Extraction()
train = self.GetTraingData()
Py_0 = 0.0
Py_1 = 0.0
Pfeature = {}
lastPfeature = {}
tmp_train = []
for line in train:
tmp_train.append(line)
tmp_feature = util_feature.GetFeature(tmp_train, self.NumOfFeature,
self.OutputFeature,
self.OutputAllFeature)
# + -
for f in tmp_feature:
Py_0 += self.E_p_base
Py_1 += self.E_p_base
Pfeature[f] = [self.E_p_base, self.E_p_base]
for line in tmp_train:
if line[0] == '0':
for col in line[1:len(line)]:
if col in Pfeature:
Py_0 += 1
Pfeature[col][1] += 1
if line[0] == '1':
for col in line[1:len(line)]:
if col in Pfeature:
Py_1 += 1
Pfeature[col][0] += 1
self.StoreModel(Py_0, Py_1, Pfeature)
print("Training all the set is down, The model is kept in the " +
self.ModelFile)
def Training(self, fold):
print("Evaluate the model by " + str(fold) +
"-fold(it will cost some time to training the model)......\n")
util_feature = FeatureExtraction.Util_Feature_Extraction()
train = self.GetTraingData()
accuracy = 0.0
percision = [0.0, 0.0]
recall = [0.0, 0.0]
F1 = [0.0, 0.0]
for i in range(fold):
Py_0 = 0.0
Py_1 = 0.0
Pfeature = {}
lastPfeature = {}
tmp_train = []
#print(len(train))
for line in train[0:int((i) * len(train) / fold)]: #
tmp_train.append(line)
for line in train[int((i + 1) * len(train) /
fold):int((fold) * len(train) / fold)]: #
tmp_train.append(line)
tmp_feature = util_feature.GetFeature(tmp_train, self.NumOfFeature,
self.OutputFeature,
self.OutputAllFeature)
# + -
for f in tmp_feature:
Py_0 += self.E_p_base
Py_1 += self.E_p_base
Pfeature[f] = [self.E_p_base, self.E_p_base]
for line in tmp_train:
if line[0] == '0':
for col in line[1:len(line)]:
if col in Pfeature:
Py_0 += 1
Pfeature[col][1] += 1
if line[0] == '1':
for col in line[1:len(line)]:
if col in Pfeature:
Py_1 += 1
Pfeature[col][0] += 1
#for f in Pfeature:
#if(Pfeature[f][0]+Pfeature[f][1] != Py_0+Py_1)
#print(Py_1+Py_0)
TP = 0.0
FP = 0.0
TN = 0.0
FN = 0.0
print(str(i + 1) + "th training is down! \n")
Evaluation_Result = []
Result = []
for line in train[int((i) * len(train) /
fold):int((i + 1) * len(train) / fold)]: #
current_P1 = math.log((Py_1 / (Py_0 + Py_1)), 10)
current_P0 = math.log((Py_0 / (Py_0 + Py_1)), 10)
for f in Pfeature:
if f in line[1:len(line)]:
current_P1 += math.log(1.25 * Pfeature[f][0] / Py_1,
10)
current_P0 += math.log(2 * Pfeature[f][1] / Py_0, 10)
#for col in line[1:len(line)-1]:
#if col in Pfeature:
#current_P1 *= Pfeature[col][0]/Py_1
#current_P0 *= Pfeature[col][1]/Py_0
if (current_P1 == 0 or current_P0 == 0):
print("equal to 0 --->error")
exit()
if current_P1 >= current_P0:
Evaluation_Result.append(1)
else:
Evaluation_Result.append(0)
if line[0] == '1':
Result.append(1)
elif line[0] == '0':
Result.append(0)
# '''
# if (current_P1 >= current_P0 and line[0] == '1'):
# TN +=1
# elif (current_P1 < current_P0 and line[0] == '0'):
# TP += 1
# elif (current_P1 >= current_P0 and line[0] == '0'):
# FN += 1
# elif (current_P1 < current_P0 and line[0] == '1'):
# FP += 1
# else:
# print("error result")
# '''
# '''
# if (current_P1 >= current_P0 and line[0] == '1'):
# TP +=1
# elif (current_P1 < current_P0 and line[0] == '0'):
# TN += 1
# elif (current_P1 >= current_P0 and line[0] == '0'):
# FP += 1
# elif (current_P1 < current_P0 and line[0] == '1'):
# FN += 1
# else:
# print("error result")
# '''
#print("Right is"+ str((TP+TN)/(TP+TN+FP+FN)))
Evaluation = precision_recall_fscore_support(Result,
Evaluation_Result,
pos_label=1)
Accuracy = accuracy_score(Result, Evaluation_Result)
#print(Evaluation)
# print("Percision OF "+str(i+1)+"th training is " + str(TP/(TP+FP)))
# print("Recall OF "+str(i+1)+"th training is " + str(TP/(TP+FN)))
# print("F1 OF "+str(i+1)+"th training is " + str(2*TP/(2*TP+FP+FN)) + '\n\n')
print("Accuracy OF " + str(i + 1) + "th training is " +
str(Accuracy))
print("Percision OF " + str(i + 1) + "th training is " +
str(Evaluation[0]))
print("Recall OF " + str(i + 1) + "th training is " +
str(Evaluation[1]))
print("F1 OF " + str(i + 1) + "th training is " +
str(Evaluation[2]) + '\n\n')
percision += Evaluation[0]
recall += Evaluation[1]
F1 += Evaluation[2]
accuracy += Accuracy
#print("by tool")
print("Now Get The Final Evaluation Of This Model:")
print("Accuracy is " + str(accuracy / 5) + "\n")
print("percision is " + str(percision / 5))
print("recall is " + str(recall / 5))
print("F1 is " + str(F1 / 5) + "\n")
