tflearn中计算混淆矩阵方法——需要经过一步转换

1def do_rnn_wordbag(trainX, testX, trainY, testY):
2    y_test=testY
3    #trainX = pad_sequences(trainX, maxlen=100, value=0.)
4    #testX = pad_sequences(testX, maxlen=100, value=0.)
5    # Converting labels to binary vectors
6    trainY = to_categorical(trainY, nb_classes=2)
7    testY = to_categorical(testY, nb_classes=2)
8  
9    # Network building
10    net = tflearn.input_data([None, 100]) 
11    net = tflearn.embedding(net, input_dim=1000, output_dim=128) 
12    net = tflearn.lstm(net, 128, dropout=0.1)
13    net = tflearn.fully_connected(net, 2, activation='softmax')
14    net = tflearn.regression(net, optimizer='adam', learning_rate=0.005,
15                             loss='categorical_crossentropy')
16  
17    # Training
18    model = tflearn.DNN(net, tensorboard_verbose=0)
19    model.fit(trainX, trainY, validation_set=0.1, show_metric=True,
20              batch_size=1,run_id="uba",n_epoch=10)
21  
22    y_predict_list = model.predict(testX)
23    #print y_predict_list
24
25    y_predict = []
26    for i in y_predict_list:
27        #print  i[0]
28        if i[0] >= 0.5:
29            y_predict.append(0)
30        else:
31            y_predict.append(1)
32
33    print(classification_report(y_test, y_predict))
34    print metrics.confusion_matrix(y_test, y_predict)
35
36    print y_train
37
38    print "ture"
39    print y_test
40    print "pre"
41    print y_predict

传统方法贝叶斯：

1def do_nb(x_train, x_test, y_train, y_test):
2    gnb = GaussianNB()
3    gnb.fit(x_train,y_train)
4    y_pred=gnb.predict(x_test)
5    print(classification_report(y_test, y_pred))
6    print metrics.confusion_matrix(y_test, y_pred)

 传统方法hmm：

1def do_hmm(trainX, testX, trainY, testY):
2    T=-580
3    N=2
4    lengths=[1]
5    X=[[0]]
6    print len(trainX)
7    for i in trainX:
8        z=[]
9        for j in i:
10            z.append([j])
11        #print z
12        #X.append(z)
13        X=np.concatenate([X,np.array(z)])
14        lengths.append(len(i))
15
16    #print lengths
17    #print X.shape
18
19
20
21    remodel = hmm.GaussianHMM(n_components=N, covariance_type="full", n_iter=100)
22    remodel.fit(X, lengths)
23
24    y_predict=[]
25    for i in testX:
26        z=[]
27        for j in i:
28            z.append([j])
29        y_pred=remodel.score(z)
30        print y_pred
31        if y_pred < T:
32            y_predict.append(1)
33        else:
34            y_predict.append(0)
35    y_predict=np.array(y_predict)
36
37    print(classification_report(testY, y_predict))
38    print metrics.confusion_matrix(testY, y_predict)
39
40    print testY
41    print y_predict