def process_svm_result(fileTest,filePredict):
#测试集的类别标签
test_y=[]
f_test=open(fileTest,'r')
for line in f_test.readlines():
lineSet=line.strip().split()
test_y.append(lineSet[0])
#预测的标签
predict=[]
f_predict=open(filePredict,'r')
for line in f_predict.readlines():
predict.append(line.strip())
#评价指标为准确率、召回率、F值
class_dict={'1':'1','2':"2"}
precision_dict=performance.calc_precision(predict,test_y,class_dict)
recall_dict=performance.calc_recall(predict,test_y,class_dict)
fscore_dict=performance.calc_fscore(predict,test_y,class_dict)
print("有情绪微博——————————————————")
print("准确率：%r"%(precision_dict['2']))
print("召回率：%r"%(recall_dict['2']))
print("F值： %r"%(fscore_dict['2']))
print("无情绪微博——————————————————")
print("准确率：%r"%(precision_dict['1']))
print("召回率：%r"%(recall_dict['1']))
print("F值： %r"%(fscore_dict['1']))
accuracy = metrics.accuracy_score(test_y, predict)
print 'accuracy: %.2f%%' % (100 * accuracy)
def evaluation(fileTest,filePredict):
#测试集的类别标签
test_y=[]
f_test=open(fileTest,'r')
for line in f_test.readlines():
lineSet=line.strip().split()
test_y.append(lineSet[2])
#预测的标签
predict=[]
f_predict=open(filePredict,'r')
for line in f_predict.readlines():
predict.append(line.strip())
#评价指标为准确率、召回率、F值
accuracy = metrics.accuracy_score(test_y, predict)
print 'accuracy: %.2f%%' % (100 * accuracy)
#输出各类指标
print("weibo情绪识别任务------------")
#宏平均
class_dict1={'happiness':'happiness','like':'like','anger':'anger',
'sadness':'sadness','fear':'fear','disgust':'disgust','surprise':'surprise'}
macro_dict1=performance.calc_macro_average(predict,test_y,class_dict1)
#微平均
micro_dict1=performance.calc_micro_average(predict,test_y,class_dict1)
#每一类情绪
class_dict={'happiness':'happiness','like':'like','anger':'anger',
'sadness':'sadness','fear':'fear','disgust':'disgust','surprise':'surprise',
'none':'none'}
#precision
precision_dict=performance.calc_precision(predict,test_y,class_dict)
print("macro_precision——%r"%(macro_dict1['macro_p']))
print("micro_precision——%r"%(micro_dict1['micro_p']))
for i in class_dict:
print("%r:%r"%(class_dict[i],precision_dict[class_dict[i]]))
#recall
recall_dict=performance.calc_recall(predict,test_y,class_dict)
print("macro_recall——%r"%(macro_dict1['macro_r']))
print("micro_recall——%r"%(micro_dict1['micro_r']))
for i in class_dict:
print("%r:%r"%(class_dict[i],recall_dict[class_dict[i]]))
#f-measure
fscore_dict=performance.calc_fscore(predict,test_y,class_dict)
print("macro_fscore——%r"%(macro_dict1['macro_f1']))
print("micro_fscore——%r"%(micro_dict1['micro_f1']))
for i in class_dict:
print("%r:%r"%(class_dict[i],fscore_dict[class_dict[i]]))
print("-------------------------")
def score(result):
#人工标注
label = []
f1 = open('weibo_label.txt', 'r')
for line in f1.readlines():
lineSet = line.strip().split()
label.append(lineSet[1])
#输出各类指标
class_dict = {'Y': 'Y'}
print("weibo情绪判断任务------------")
precision_dict = performance.calc_precision(result, label, class_dict)
print("precision:%r" % (precision_dict['Y']))
recall_dict = performance.calc_recall(result, label, class_dict)
print("recall:%r" % (recall_dict['Y']))
fscore_dict = performance.calc_fscore(result, label, class_dict)
print("f1:%r" % (fscore_dict['Y']))
print("-------------------------")
def score(result):
#人工标注
label=[]
f1=open('label.txt','r')
for line in f1.readlines():
lineSet=line.strip().split()
label.append(lineSet[1])
#输出各类指标
class_dict={'Y':'Y'}
print("weibo情绪判断任务------------")
precision_dict=performance.calc_precision(result,label,class_dict)
print("precision:%r"%(precision_dict['Y']) )
recall_dict=performance.calc_recall(result,label,class_dict)
print("recall:%r"%(recall_dict['Y']) )
fscore_dict=performance.calc_fscore(result,label,class_dict)
print("f1:%r"%(fscore_dict['Y']))
print("-------------------------")
def merge_svm_GBDT():
merge_result = []
threshold_minus = 1.496
threshold_plus = 1.65
#svm的结果
svm_predict = []
f_svm = open('temporary/svm_predict', 'r')
for line in f_svm.readlines():
svm_predict.append(line.strip())
#GBDT的结果
f_GBDT = open('temporary/GBDT_predict1000', 'r')
index = 0
for line in f_GBDT.readlines():
num = float(line.strip())
if num < threshold_minus:
merge_result.append('1')
elif num > threshold_plus:
merge_result.append('2')
else:
merge_result.append(svm_predict[index])
index += 1
#测试集的类别标签
test_y = []
f_test = open('temporary/test.sample', 'r')
for line in f_test.readlines():
lineSet = line.strip().split()
test_y.append(lineSet[0])
#评价指标为准确率、召回率、F值
class_dict = {'1': '1', '2': "2"}
precision_dict = performance.calc_precision(merge_result, test_y,
class_dict)
recall_dict = performance.calc_recall(merge_result, test_y, class_dict)
fscore_dict = performance.calc_fscore(merge_result, test_y, class_dict)
print("有情绪微博——————————————————")
print("准确率：%r" % (precision_dict['2']))
print("召回率：%r" % (recall_dict['2']))
print("F值： %r" % (fscore_dict['2']))
print("无情绪微博——————————————————")
print("准确率：%r" % (precision_dict['1']))
print("召回率：%r" % (recall_dict['1']))
print("F值： %r" % (fscore_dict['1']))
accuracy = metrics.accuracy_score(test_y, merge_result)
print 'accuracy: %.2f%%' % (100 * accuracy)
def score_emotion(fileName1,fileName2):
#标注结果
label=[]
f=open(fileName1,'r')
for line in f.readlines():
labelLine=line.strip().split()
label.append(labelLine[2])
#预测结果
result=[]
f1=open(fileName2,'r')
for line in f1.readlines():
result.append(line.strip())
#输出各类指标
print("weibo情绪识别任务------------")
#宏平均
class_dict1={'happiness':'happiness','like':'like','anger':'anger',
'sadness':'sadness','fear':'fear','disgust':'disgust','surprise':'surprise'}
macro_dict1=performance.calc_macro_average(result,label,class_dict1)
#每一类情绪
class_dict={'happiness':'happiness','like':'like','anger':'anger',
'sadness':'sadness','fear':'fear','disgust':'disgust','surprise':'surprise',
'none':'none'}
#precision
precision_dict=performance.calc_precision(result,label,class_dict)
print("macro_precision——%r"%(macro_dict1['macro_p']))
for i in class_dict:
print("%r:%r"%(class_dict[i],precision_dict[class_dict[i]]))
#recall
recall_dict=performance.calc_recall(result,label,class_dict)
print("macro_recall——%r"%(macro_dict1['macro_r']))
for i in class_dict:
print("%r:%r"%(class_dict[i],recall_dict[class_dict[i]]))
#f-measure
fscore_dict=performance.calc_fscore(result,label,class_dict)
print("macro_fscore——%r"%(macro_dict1['macro_f1']))
for i in class_dict:
print("%r:%r"%(class_dict[i],fscore_dict[class_dict[i]]))
print("-------------------------")
def merge_svm_GBDT():
merge_result=[]
threshold_minus=1.496
threshold_plus=1.65
#svm的结果
svm_predict=[]
f_svm=open('temporary/svm_predict','r')
for line in f_svm.readlines():
svm_predict.append(line.strip())
#GBDT的结果
f_GBDT=open('temporary/GBDT_predict1000','r')
index=0
for line in f_GBDT.readlines():
num=float(line.strip())
if num < threshold_minus:
merge_result.append('1')
elif num > threshold_plus:
merge_result.append('2')
else:
merge_result.append(svm_predict[index])
index+=1
#测试集的类别标签
test_y=[]
f_test=open('temporary/test.sample','r')
for line in f_test.readlines():
lineSet=line.strip().split()
test_y.append(lineSet[0])
#评价指标为准确率、召回率、F值
class_dict={'1':'1','2':"2"}
precision_dict=performance.calc_precision(merge_result,test_y,class_dict)
recall_dict=performance.calc_recall(merge_result,test_y,class_dict)
fscore_dict=performance.calc_fscore(merge_result,test_y,class_dict)
print("有情绪微博——————————————————")
print("准确率：%r"%(precision_dict['2']))
print("召回率：%r"%(recall_dict['2']))
print("F值： %r"%(fscore_dict['2']))
print("无情绪微博——————————————————")
print("准确率：%r"%(precision_dict['1']))
print("召回率：%r"%(recall_dict['1']))
print("F值： %r"%(fscore_dict['1']))
accuracy = metrics.accuracy_score(test_y, merge_result)
print 'accuracy: %.2f%%' % (100 * accuracy)
def score_emotion(fileName1,fileName2):
#标注结果
label=[]
f=open(fileName1,'r')
for line in f.readlines():
labelLine=line.strip().split()
label.append(labelLine[2])
#预测结果
result=[]
f1=open(fileName2,'r')
for line in f1.readlines():
result.append(line.strip())
#输出各类指标
print("weibo情绪识别任务------------")
#宏平均
class_dict1={'happiness':'happiness','like':'like','anger':'anger',
'sadness':'sadness','fear':'fear','disgust':'disgust','surprise':'surprise'}
macro_dict1=performance.calc_macro_average(result,label,class_dict1)
#每一类情绪
class_dict={'happiness':'happiness','like':'like','anger':'anger',
'sadness':'sadness','fear':'fear','disgust':'disgust','surprise':'surprise',
'none':'none'}
#precision
precision_dict=performance.calc_precision(result,label,class_dict)
print("macro_precision——%r"%(macro_dict1['macro_p']))
for i in class_dict:
print("%r:%r"%(class_dict[i],precision_dict[class_dict[i]]))
#recall
recall_dict=performance.calc_recall(result,label,class_dict)
print("macro_recall——%r"%(macro_dict1['macro_r']))
for i in class_dict:
print("%r:%r"%(class_dict[i],recall_dict[class_dict[i]]))
#f-measure
fscore_dict=performance.calc_fscore(result,label,class_dict)
print("macro_fscore——%r"%(macro_dict1['macro_f1']))
for i in class_dict:
print("%r:%r"%(class_dict[i],fscore_dict[class_dict[i]]))
print("-------------------------")
'surprise': 'surprise'
}
macro_dict1 = performance.calc_macro_average(result, label, class_dict1)
#每一类情绪
class_dict = {
'happiness': 'happiness',
'like': 'like',
'anger': 'anger',
'sadness': 'sadness',
'fear': 'fear',
'disgust': 'disgust',
'surprise': 'surprise',
'none': 'none'
}
#precision
precision_dict = performance.calc_precision(result, label, class_dict)
print("macro_precision——%r" % (macro_dict1['macro_p']))
for i in class_dict:
print("%r:%r" % (class_dict[i], precision_dict[class_dict[i]]))
#recall
recall_dict = performance.calc_recall(result, label, class_dict)
print("macro_recall——%r" % (macro_dict1['macro_r']))
for i in class_dict:
print("%r:%r" % (class_dict[i], recall_dict[class_dict[i]]))
#f-measure
fscore_dict = performance.calc_fscore(result, label, class_dict)
print("macro_fscore——%r" % (macro_dict1['macro_f1']))
for i in class_dict:
print("%r:%r" % (class_dict[i], fscore_dict[class_dict[i]]))
print("-------------------------")
# label=readin_label('weibo_label.txt')
#正确率acc
accuracy=performance.calc_acc(result,label)
print("正确率：%r"%(accuracy))
#宏平均
class_dict1={'happiness':'happiness','like':'like','anger':'anger',
'sadness':'sadness','fear':'fear','disgust':'disgust','surprise':'surprise'}
macro_dict1=performance.calc_macro_average(result,label,class_dict1)
#每一类情绪
class_dict={'happiness':'happiness','like':'like','anger':'anger',
'sadness':'sadness','fear':'fear','disgust':'disgust','surprise':'surprise',
'none':'none'}
#precision
precision_dict=performance.calc_precision(result,label,class_dict)
print("macro_precision——%r"%(macro_dict1['macro_p']))
for i in class_dict:
print("%r:%r"%(class_dict[i],precision_dict[class_dict[i]]))
#recall
recall_dict=performance.calc_recall(result,label,class_dict)
print("macro_recall——%r"%(macro_dict1['macro_r']))
for i in class_dict:
print("%r:%r"%(class_dict[i],recall_dict[class_dict[i]]))
#f-measure
fscore_dict=performance.calc_fscore(result,label,class_dict)
print("macro_fscore——%r"%(macro_dict1['macro_f1']))
for i in class_dict:
print("%r:%r"%(class_dict[i],fscore_dict[class_dict[i]]))
print("-------------------------")
def gradient_boosting_hhh(fname_samp_train, fname_samp_test,len_term_set):
#for_train_x每一行是词典，train_y是list
for_train_x,train_y=pytc.load_samps(fname_samp_train, fs_num=0)
for_test_x,test_y=pytc.load_samps(fname_samp_test, fs_num=0)
train_x=process_for_train_x_or_test_x(for_train_x,len_term_set)
test_x=process_for_train_x_or_test_x(for_test_x,len_term_set)
#回归
# train_x = np.float32(train_x)
# train_y = np.float32(train_y)
# test_x = np.float32(test_x)
# test_y = np.float32(test_y)
num_train,num_feat = np.shape(train_x)
num_test,num_feat = np.shape(test_x)
print '******************** 数据信息 *********************'
print '#训练集数据: %d, #测试集数据: %d, 维度: %d' % (num_train, num_test, num_feat)
start_time = time.time()
#分类
model = gradient_boosting_classifier(train_x, train_y)
#回归
# model = gradient_boosting_regressor(train_x, train_y)
print '训练耗时 %fs!' % (time.time() - start_time)
predict = model.predict(test_x)
#输出预测标签
fw_predict=open('predict_classifier','w')
for item in predict:
print>>fw_predict,item
# #设置阈值
# threshold= 1.539
# f_w=open('temporary/GBDT_predict1000','w')
# f_w_label=open('temporary/GBDT_predict_label1000','w')
# for item in predict:
# print >>f_w,item
# if item>=threshold:
# item=2
# else:
# item=1
# print>>f_w_label,item
#测试集的类别标签
test_y=[]
f_test=open('temporary/test.sample','r')
for line in f_test.readlines():
lineSet=line.strip().split()
test_y.append(lineSet[0])
# #预测的标签
# predict=[]
# f_predict=open('temporary/GBDT_predict','r')
# for line in f_predict.readlines():
# predict.append(line.strip())
#评价指标为准确率、召回率、F值
class_dict={'1':'1','2':"2"}
precision_dict=performance.calc_precision(predict,test_y,class_dict)
recall_dict=performance.calc_recall(predict,test_y,class_dict)
fscore_dict=performance.calc_fscore(predict,test_y,class_dict)
print("有情绪微博——————————————————")
print("准确率：%r"%(precision_dict['2']))
print("召回率：%r"%(recall_dict['2']))
print("F值： %r"%(fscore_dict['2']))
print("无情绪微博——————————————————")
print("准确率：%r"%(precision_dict['1']))
print("召回率：%r"%(recall_dict['1']))
print("F值： %r"%(fscore_dict['1']))
accuracy = metrics.accuracy_score(test_y, predict)
print 'accuracy: %.2f%%' % (100 * accuracy)
