BIDI_ML_INFO_EXTRACTION/BiddingKG/dl/test/test_model_fjs.py

import sys

from keras.models import Model
from keras.layers import Input, LSTM, Dense
import numpy as np
import pandas as pd
from matplotlib import pyplot

from BiddingKG.dl.common.models import *
from sklearn.metrics import classification_report


sys.path.append(os.path.abspath("../.."))
model_file = "model_person_classify_fjs.model.hdf5"



def getSeq2seqModel():
    # Batch size for training.
    batch_size = 64
    # Number of epochs to train for.
    epochs = 100
    # Latent dimensionality of the encoding space.
    latent_dim = 256
    # Number of samples to train on.
    num_samples = 10000
    # Path to the data txt file on disk.
    data_path = 'fra-eng/fra.txt'

    # Vectorize the data.
    input_texts = []
    target_texts = []
    # Set方便去重
    input_characters = set()
    target_characters = set()
    with open(data_path, 'r', encoding='utf-8') as f:
        lines = f.read().split('\n')
    for line in lines[: min(num_samples, len(lines) - 1)]:
        input_text, target_text, _ = line.split('\t')
        # 句子开始符：\t  句子终止符：\n
        # We use "tab" as the "start sequence" character
        # for the targets, and "\n" as "end sequence" character.
        target_text = '\t' + target_text + '\n'
        input_texts.append(input_text)
        target_texts.append(target_text)
        for char in input_text:
            if char not in input_characters:
                input_characters.add(char)
        for char in target_text:
            if char not in target_characters:
                target_characters.add(char)

    # 将字符排序
    input_characters = sorted(list(input_characters))
    target_characters = sorted(list(target_characters))
    # Encoder的输入类别长度：字符表长度
    # Decoder的输出类别长度：字符表长度
    num_encoder_tokens = len(input_characters)
    num_decoder_tokens = len(target_characters)
    # Encoder的输入最大长度：最长的句子的长度
    # Decoder的输入最大长度：最长的句子的长度
    max_encoder_seq_length = max([len(txt) for txt in input_texts])
    max_decoder_seq_length = max([len(txt) for txt in target_texts])

    print('Number of samples:', len(input_texts))
    print('Number of unique input tokens:', num_encoder_tokens)
    print('Number of unique output tokens:', num_decoder_tokens)
    print('Max sequence length for inputs:', max_encoder_seq_length)
    print('Max sequence length for outputs:', max_decoder_seq_length)

    # 为每个字符与下标组成一个字典
    input_token_index = dict(
        [(char, i) for i, char in enumerate(input_characters)])
    target_token_index = dict(
        [(char, i) for i, char in enumerate(target_characters)])

    # 初始化Encoder输入矩阵
    # 第一维为句子条数，即RNN循环次数
    # 第二维为Encoder最大输入长度
    # 第三维为Encoder的输入类别长度
    encoder_input_data = np.zeros(
        (len(input_texts), max_encoder_seq_length, num_encoder_tokens),
        dtype='float32')

    # 初始化Decoder输入矩阵（Encoder输出）
    # 第一维为句子条数，即RNN循环次数
    # 第二维为Decoder最大输入长度
    # 第三维为Decoder的输入类别长度
    decoder_input_data = np.zeros(
        (len(input_texts), max_decoder_seq_length, num_decoder_tokens),
        dtype='float32')

    # 初始化Decoder输出矩阵
    # 第一维为句子条数，即RNN循环次数
    # 第二维为Decoder最大输入长度
    # 第三维为Decoder的输入类别长度
    decoder_target_data = np.zeros(
        (len(input_texts), max_decoder_seq_length, num_decoder_tokens),
        dtype='float32')

    # 将input和target打包成一对，如[input, target]
    for i, (input_text, target_text) in enumerate(zip(input_texts, target_texts)):

        for t, char in enumerate(input_text):
            encoder_input_data[i, t, input_token_index[char]] = 1.
        encoder_input_data[i, t + 1:, input_token_index[' ']] = 1.

        for t, char in enumerate(target_text):
            # decoder_target_data is ahead of decoder_input_data by one timestep
            decoder_input_data[i, t, target_token_index[char]] = 1.
            if t > 0:
                # decoder_target_data will be ahead by one timestep
                # and will not include the start character.
                decoder_target_data[i, t - 1, target_token_index[char]] = 1.
        decoder_input_data[i, t + 1:, target_token_index[' ']] = 1.
        decoder_target_data[i, t:, target_token_index[' ']] = 1.

    # Define an input sequence and process it.
    encoder_inputs = Input(shape=(None, num_encoder_tokens))
    encoder = LSTM(latent_dim, return_state=True)
    encoder_outputs, state_h, state_c = encoder(encoder_inputs)
    # We discard `encoder_outputs` and only keep the states.
    encoder_states = [state_h, state_c]

    # Set up the decoder, using `encoder_states` as initial state.
    decoder_inputs = Input(shape=(None, num_decoder_tokens))
    # We set up our decoder to return full output sequences,
    # and to return internal states as well. We don't use the
    # return states in the training model, but we will use them in inference.
    decoder_lstm = LSTM(latent_dim, return_sequences=True, return_state=True)
    decoder_outputs, _, _ = decoder_lstm(decoder_inputs,
                                         initial_state=encoder_states)
    decoder_dense = Dense(num_decoder_tokens, activation='softmax')
    decoder_outputs = decoder_dense(decoder_outputs)

    # Define the model that will turn
    # `encoder_input_data` & `decoder_input_data` into `decoder_target_data`
    model = Model([encoder_inputs, decoder_inputs], decoder_outputs)

    # Run training
    model.compile(optimizer='rmsprop', loss='categorical_crossentropy',
                  metrics=['accuracy'])
    model.fit([encoder_input_data, decoder_input_data], decoder_target_data,
              batch_size=batch_size,
              epochs=epochs,
              validation_split=0.2)
    # Save model
    model.save('s2s.h5')

    # Next: inference mode (sampling).
    # Here's the drill:
    # 1) encode input and retrieve initial decoder state
    # 2) run one step of decoder with this initial state
    # and a "start of sequence" token as target.
    # Output will be the next target token
    # 3) Repeat with the current target token and current states

    # Define sampling models
    encoder_model = Model(encoder_inputs, encoder_states)

    decoder_state_input_h = Input(shape=(latent_dim,))
    decoder_state_input_c = Input(shape=(latent_dim,))
    decoder_states_inputs = [decoder_state_input_h, decoder_state_input_c]
    decoder_outputs, state_h, state_c = decoder_lstm(
        decoder_inputs, initial_state=decoder_states_inputs)
    decoder_states = [state_h, state_c]
    decoder_outputs = decoder_dense(decoder_outputs)
    decoder_model = Model(
        [decoder_inputs] + decoder_states_inputs,
        [decoder_outputs] + decoder_states)

    # Reverse-lookup token index to decode sequences back to
    # something readable.
    reverse_input_char_index = dict(
        (i, char) for char, i in input_token_index.items())
    reverse_target_char_index = dict(
        (i, char) for char, i in target_token_index.items())

    def decode_sequence(input_seq):
        # Encode the input as state vectors.
        states_value = encoder_model.predict(input_seq)

        # Generate empty target sequence of length 1.
        target_seq = np.zeros((1, 1, num_decoder_tokens))
        # Populate the first character of target sequence with the start character.
        target_seq[0, 0, target_token_index['\t']] = 1.

        # Sampling loop for a batch of sequences
        # (to simplify, here we assume a batch of size 1).
        stop_condition = False
        decoded_sentence = ''
        while not stop_condition:
            output_tokens, h, c = decoder_model.predict(
                [target_seq] + states_value)

            # Sample a token
            sampled_token_index = np.argmax(output_tokens[0, -1, :])
            sampled_char = reverse_target_char_index[sampled_token_index]
            decoded_sentence += sampled_char

            # Exit condition: either hit max length
            # or find stop character.
            if (sampled_char == '\n' or
                    len(decoded_sentence) > max_decoder_seq_length):
                stop_condition = True

            # Update the target sequence (of length 1).
            target_seq = np.zeros((1, 1, num_decoder_tokens))
            target_seq[0, 0, sampled_token_index] = 1.

            # Update states
            states_value = [h, c]

        return decoded_sentence


    for seq_index in range(100):
        # Take one sequence (part of the training set)
        # for trying out decoding.
        input_seq = encoder_input_data[seq_index: seq_index + 1]
        decoded_sentence = decode_sequence(input_seq)
        print('-')
        print('Input sentence:', input_texts[seq_index])
        print('Decoded sentence:', decoded_sentence)


def getBiLSTM_Dropout():
    '''
    @summary: 获得模型
    '''
    input_shape = (2, 35, 128)
    # input_shape = (1, 70, 128)
    output_shape = [5]

    L_input = layers.Input(shape=input_shape[1:], dtype="float32")
    R_input = layers.Input(shape=input_shape[1:], dtype="float32")
    lstm_0 = layers.Bidirectional(layers.LSTM(32, dropout=0.5, recurrent_dropout=0.5, return_sequences=True))(L_input)
    avg_0 = layers.GlobalAveragePooling1D()(lstm_0)
    lstm_2 = layers.Bidirectional(layers.LSTM(32, dropout=0.5, recurrent_dropout=0.5, return_sequences=True))(R_input)
    avg_2 = layers.GlobalAveragePooling1D()(lstm_2)
    concat = layers.merge([avg_0, avg_2], mode="concat")
    output = layers.Dense(output_shape[0], activation="softmax")(concat)

    model = models.Model(inputs=[L_input, R_input], outputs=output)
    model.compile(optimizer=optimizers.Adam(lr=0.0005), loss=losses.binary_crossentropy, metrics=[precision, recall, f1_score])
    return model


def getBiRNN_Dropout():
    '''
    @summary: 获得模型
    '''
    input_shape = (2, 10, 128)
    output_shape = [5]

    L_input = layers.Input(shape=input_shape[1:], dtype="float32")
    R_input = layers.Input(shape=input_shape[1:], dtype="float32")
    lstm_0 = layers.Bidirectional(layers.SimpleRNN(32, dropout=0.65, recurrent_dropout=0.65, return_sequences=True))(L_input)
    avg_0 = layers.GlobalAveragePooling1D()(lstm_0)
    lstm_2 = layers.Bidirectional(layers.SimpleRNN(32, dropout=0.65, recurrent_dropout=0.65, return_sequences=True))(R_input)
    avg_2 = layers.GlobalAveragePooling1D()(lstm_2)
    concat = layers.merge([avg_0, avg_2], mode="concat")
    output = layers.Dense(output_shape[0], activation="softmax")(concat)

    model = models.Model(inputs=[L_input, R_input], outputs=output)
    model.compile(optimizer=optimizers.Adam(lr=0.0005), loss=losses.binary_crossentropy, metrics=[precision, recall, f1_score])
    return model


def getBiGRU_Dropout():
    '''
    @summary: 获得模型
    '''
    input_shape = (2, 35, 128)
    # input_shape = (1, 70, 128)
    output_shape = [5]

    L_input = layers.Input(shape=input_shape[1:], dtype="float32")
    R_input = layers.Input(shape=input_shape[1:], dtype="float32")
    lstm_0 = layers.Bidirectional(layers.GRU(32, dropout=0.4, recurrent_dropout=0.4, return_sequences=True))(L_input)
    avg_0 = layers.GlobalAveragePooling1D()(lstm_0)
    lstm_2 = layers.Bidirectional(layers.GRU(32, dropout=0.4, recurrent_dropout=0.4, return_sequences=True))(R_input)
    avg_2 = layers.GlobalAveragePooling1D()(lstm_2)
    concat = layers.merge([avg_0, avg_2], mode="concat")
    output = layers.Dense(output_shape[0], activation="softmax")(concat)

    model = models.Model(inputs=[L_input, R_input], outputs=output)
    model.compile(optimizer=optimizers.Adam(lr=0.0005), loss=losses.binary_crossentropy, metrics=[precision, recall, f1_score])
    return model


def getLSTM_Dropout():
    '''
    @summary: 获得模型
    '''
    input_shape = (2, 10, 128)
    output_shape = [5]

    input = layers.Input(shape=input_shape[1:], dtype="float32")
    lstm = layers.LSTM(32, dropout=0.2, recurrent_dropout=0.2, return_sequences=True)(input)
    avg = layers.GlobalAveragePooling1D()(lstm)
    output = layers.Dense(output_shape[0], activation="softmax")(avg)

    model = models.Model(inputs=input, outputs=output)
    model.compile(optimizer=optimizers.Adam(lr=0.0005), loss=losses.binary_crossentropy, metrics=[precision, recall, f1_score])
    return model


def getGRUModel_Dropout():
    '''
    @summary: 获得模型
    '''
    # input_shape = (2, 10, 128)
    input_shape = (1, 70, 128)
    output_shape = [5]

    input = layers.Input(shape=input_shape[1:], dtype="float32")
    gru = layers.GRU(32, dropout=0.15, recurrent_dropout=0.15, return_sequences=True)(input)
    avg = layers.GlobalAveragePooling1D()(gru)
    output = layers.Dense(output_shape[0], activation="softmax")(avg)

    model = models.Model(inputs=input, outputs=output)
    model.compile(optimizer=optimizers.Adam(lr=0.0005), loss=losses.binary_crossentropy, metrics=[precision, recall, f1_score])
    return model


def getRNNModel_Dropout():
    '''
    @summary: 获得模型
    '''
    input_shape = (2, 10, 128)
    output_shape = [5]

    input = layers.Input(shape=input_shape[1:], dtype="float32")
    rnn = layers.SimpleRNN(32, dropout=0.5, recurrent_dropout=0.5, return_sequences=True)(input)
    avg = layers.GlobalAveragePooling1D()(rnn)
    output = layers.Dense(output_shape[0], activation="softmax")(avg)

    model = models.Model(inputs=input, outputs=output)
    model.compile(optimizer=optimizers.Adam(lr=0.0005), loss=losses.binary_crossentropy, metrics=[precision, recall, f1_score])
    return model


def getGCNModel():
    return


def getData3(isTrain = True):

    '''
    :return:返回训练数据或测试数据的词嵌入，分前后两个句子，不包含中心词
    '''
    df = pd.read_csv("C:\\Users\\admin\\Desktop\\Person_Sentence_Notest.csv")
    df1 = pd.read_csv("C:\\Users\\admin\\Desktop\\test2000.csv")

    test_data_len = df.shape[0] * 0.2
    if isTrain:
        test_data_len = 0
    else:
        test_data_len = 3700
        df = df1

    df = df.reset_index()

    input_shape = (2, 35, 128)
    output_shape = [5]
    allLimit = 250000
    all = 0
    data_x = []
    data_y = []
    data_context = []
    for index, row in df.iterrows():
        if isTrain:
            if index < test_data_len:
                continue
        else:
            if index >= test_data_len:
                break
        if all >= allLimit:
            break

        tokens_list_front = []
        tokens_list_behind = []
        tokens_list_all = []
        sss = row["Sentence"].split("||")
        front = sss[0]
        behind = sss[2]

        ss_front = front.split(" ")
        ss_behind = behind.split(" ")
        for s in ss_front:
            tokens_list_front.append(s)
        for s in ss_behind:
            tokens_list_behind.append(s)
        tokens_list_all.append(tokens_list_front)
        tokens_list_all.append(tokens_list_behind)

        # print(np.array(tokens_list_all).shape)
        item_x = embedding(tokens_list_all, shape=input_shape)
        item_y = np.zeros(output_shape)
        item_y[row[3]] = 1
        all += 1
        data_x.append(item_x)
        data_y.append(item_y)

    print(np.array(data_x).shape, np.array(data_y).shape)
    return np.transpose(np.array(data_x), (1, 0, 2, 3)), np.array(data_y), data_context


def getData2(isTrain = True):
    '''
    :return:返回训练数据或测试数据的词嵌入，前后连成一个句子，包含中心词
    '''
    df = pd.read_csv("C:\\Users\\admin\\Desktop\\Person_Sentence_Notest.csv")
    df1 = pd.read_csv("C:\\Users\\admin\\Desktop\\test2000.csv")

    test_data_len = df.shape[0] * 0.2
    if isTrain:
        test_data_len = 0
    else:
        test_data_len = 3700
        df = df1

    df = df.reset_index()

    input_shape = (1, 70, 128)
    output_shape = [5]
    allLimit = 250000
    all = 0
    data_x = []
    data_y = []
    data_context = []
    for index, row in df.iterrows():
        if isTrain:
            if index < test_data_len:
                continue
        else:
            if index >= test_data_len:
                break
        if all >= allLimit:
            break

        tokens_list = []
        tokens_list_all = []
        ss = row["Sentence"].split(" ")
        for s in ss:
            tokens_list.append(s)
        tokens_list_all.append(tokens_list)

        item_x = embedding(tokens_list_all, shape=input_shape)
        item_y = np.zeros(output_shape)
        item_y[row[3]] = 1
        all += 1
        data_x.append(item_x)
        data_y.append(item_y)

    print(np.array(data_x).shape, np.array(data_y).shape)
    return np.transpose(np.array(data_x), (1, 0, 2, 3)), np.array(data_y), data_context


def getData(isTrain = True):
    '''

    :return:返回训练数据或测试数据的词嵌入
    '''
    df = pd.read_csv("C:\\Users\\admin\\Desktop\\Person_Sentence_Notest.csv")
    df1 = pd.read_csv("C:\\Users\\admin\\Desktop\\test2000.csv")

    test_data_len = df.shape[0] * 0.2
    if isTrain:
        test_data_len = 0
    else:
        test_data_len = 3700
        df = df1

    df = df.reset_index()

    input_shape = (2, 35, 128)
    output_shape = [5]
    allLimit = 250000
    all = 0
    data_x = []
    data_y = []
    data_context = []
    for index, row in df.iterrows():
        if isTrain:
            if index < test_data_len:
                continue
        else:
            if index >= test_data_len:
                break
        if all >= allLimit:
            break

        print(np.array(spanWindow(tokens=row["Sentence"], begin_index=row["begin_index"], end_index=row["end_index"], size=input_shape[1])).shape)

        item_x = embedding(spanWindow(tokens=row["Sentence"], begin_index=row["begin_index"], end_index=row["end_index"], size=input_shape[1]), shape=input_shape)
        item_y = np.zeros(output_shape)
        item_y[row[3]] = 1
        all += 1
        data_x.append(item_x)
        data_y.append(item_y)

    print(np.array(data_x).shape, np.array(data_y).shape)
    # print(data_x, data_y, data_context)
    return np.transpose(np.array(data_x), (1, 0, 2, 3)), np.array(data_y), data_context


def train():
    '''
    @summary: 训练模型
    '''
    model = getBiGRU_Dropout()
    model.summary()
    train_x, train_y, _ = getData3(isTrain=True)
    test_x, test_y, test_context = getData3(isTrain=False)

    # 回调checkpoint，保存loss最小的模型
    checkpoint = ModelCheckpoint(model_file, monitor="val_loss", verbose=1, save_best_only=True, mode='min')
    history_model = model.fit(x=[train_x[0], train_x[1]], y=train_y, validation_data=([test_x[0], test_x[1]], test_y), epochs=200, batch_size=256, shuffle=True, callbacks=[checkpoint])
    # history_model = model.fit(x=[train_x[0], train_x[0]], y=train_y, validation_data=([test_x[0], test_x[0]], test_y), class_weight='auto', epochs=100, batch_size=256, shuffle=True, callbacks=[checkpoint])
    # history_model = model.fit(x=[train_x[0], train_x[0]], y=train_y, validation_split=0.2, class_weight='auto', epochs=200, batch_size=256, shuffle=True, callbacks=[checkpoint])

    # 单向模型
    # history_model = model.fit(x=train_x[0], y=train_y, validation_data=([test_x[0], test_y]), class_weight='auto', epochs=200, batch_size=256, shuffle=True, callbacks=[checkpoint])
    # history_model = model.fit(x=train_x[0], y=train_y, validation_split=0.2, class_weight='auto', epochs=200, batch_size=256, shuffle=True, callbacks=[checkpoint])

    # history_model = model.fit(x=[train_x[0], train_x[1]], y=train_y, validation_split=0.2, epochs=100, class_weight='auto', batch_size=256, shuffle=True, callbacks=[checkpoint])
    # history_model = model.fit(x=[train_x[0], train_x[1]], y=train_y, validation_split=0.2, epochs=250, batch_size=256, shuffle=True, callbacks=[checkpoint])
    plotTrainTestLoss(history_model)


def predict():
    model = models.load_model(model_file, custom_objects={'precision': precision, 'recall': recall, 'f1_score': f1_score})
    test_x, test_y, test_context = getData3(isTrain=False)
    predict_y = model.predict([test_x[0], test_x[1]])
    # predict_y = model.predict([test_x[0], test_x[0]])
    # predict_y = model.predict([test_x[0]])
    targets_name = ['人名', '联系人', '招标联系人', '代理联系人', '评审专家']
    print(classification_report(np.argmax(test_y, axis=1), np.argmax(predict_y, axis=1), target_names=targets_name))
    return predict_y


def predict2Csv():
    df = pd.DataFrame(np.argmax(predict(), axis=1))
    df1 = pd.read_csv("C:\\Users\\admin\\Desktop\\test2000.csv")

    df1 = df1[0:3700]
    df1["predict_Label"] = df

    df1.to_csv("C:\\Users\\admin\\Desktop\\result3.csv")


def plotTrainTestLoss(history_model):
    pyplot.plot(history_model.history['loss'])
    pyplot.plot(history_model.history['val_loss'])
    pyplot.title('model train vs validation loss')
    pyplot.ylabel('loss')
    pyplot.xlabel('epoch')
    pyplot.legend(['train', 'validation'], loc='upper right')
    pyplot.show()


if __name__ == "__main__":
    # getData()
    train()
    predict()
    # predict2Csv()

    # getData3()