利⽤CNN进⾏多分类的⽂档分类# coding: utf-8
import tensorflow as tf
class TCNNConfig(object):有丰胸的胸罩吗
"""CNN配置参数"""
embedding_dim = 20 # 词向量维度
seq_length = 100 # 序列长度
num_classes = 73 # 类别数
num_filters = 256 # 卷积核数⽬
kernel_size = 5 # 卷积核尺⼨
vocab_size = 5000 # 词汇表达⼩
hidden_dim = 128 # 全连接层神经元
dropout_keep_prob = 0.8 # dropout保留⽐例
learning_rate = 0.001 # 学习率
batch_size = 128 # 每批训练⼤⼩
num_epochs = 5 # 总迭代轮次
print_per_batch = 100 # 每多少轮输出⼀次结果
save_per_batch = 10 # 每多少轮存⼊tensorboard
class TextCNN(object):
"""⽂本分类，CNN模型"""
def__init__(self, config):
# 三个待输⼊的数据
self.input_x = tf.placeholder(tf.int32, [None, fig.seq_length], name='input_x')
self.input_y = tf.placeholder(tf.float32, [None, fig.num_classes], name='input_y')
self.keep_prob = tf.placeholder(tf.float32, name='keep_prob')
<()
def cnn(self):
"""CNN模型"""
# 词向量映射
with tf.device('/cpu:0'):
embedding = tf.get_variable('embedding', [fig.vocab_size, bedding_dim])
embedding_inputs = bedding_lookup(embedding, self.input_x)
with tf.name_scope("cnn"):
# CNN layer
conv = v1d(embedding_inputs, fig.num_filters, fig.kernel_size, name='conv') # global max pooling layer
gmp = tf.reduce_max(conv, reduction_indices=[1], name='gmp')
with tf.name_scope("score"):
# 全连接层，后⾯接dropout以及relu激活
fc = tf.layers.dense(gmp, fig.hidden_dim, name='fc1')
fc = tf.contrib.layers.dropout(fc, self.keep_prob)
fc = lu(fc)
自制鱼缸灯架
# 分类器
self.logits = tf.layers.dense(fc, fig.num_classes, name='fc2')
self.y_pred_cls = tf.softmax(self.logits), 1) # 预测类别
with tf.name_scope("optimize"):
# 损失函数，交叉熵
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=self.logits, labels=self.input_y)
self.loss = tf.reduce_mean(cross_entropy)
# 优化器
self.optim = tf.train.AdamOptimizer(learning_fig.learning_rate).minimize(self.loss)
with tf.name_scope("accuracy"):
# 准确率
correct_pred = tf.equal(tf.argmax(self.input_y, 1), self.y_pred_cls)
self.acc = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
#!/usr/bin/python
# -*- coding: utf-8 -*-
from__future__import print_function
import os
import sys
import time
from datetime import timedelta
import numpy as np
import tensorflow as tf
from sklearn import metrics
from cnn_model import TCNNConfig, TextCNN
ws_loader import read_vocab, read_category, batch_iter, process_file, build_vocab base_dir = 'data/'
train_dir = os.path.join(base_dir, '')
test_dir = os.path.join(base_dir, '')
val_dir = os.path.join(base_dir, '')
vocab_dir = os.path.join(base_dir, '')
save_dir = 'checkpoints/textcnn'
save_path = os.path.join(save_dir, 'best_validation') # 最佳验证结果保存路径 def get_time_dif(start_time):
"""获取已使⽤时间"""
end_time = time.time()
time_dif = end_time - start_time
return timedelta(seconds=int(round(time_dif)))
def feed_data(x_batch, y_batch, keep_prob):
feed_dict = {
model.input_x: x_batch,
MOSFET裸片
model.input_y: y_batch,
model.keep_prob: keep_prob
}
return feed_dict
def evaluate(sess, x_, y_):
"""评估在某⼀数据上的准确率和损失"""
data_len = len(x_)
batch_eval = batch_iter(x_, y_, 128)
total_loss = 0.0
total_acc = 0.0
for x_batch, y_batch in batch_eval:
batch_len = len(x_batch)
feed_dict = feed_data(x_batch, y_batch, 1.0)
loss, acc = sess.run([model.loss, model.acc], feed_dict=feed_dict)
total_loss += loss * batch_len
total_acc += acc * batch_len
return total_loss / data_len, total_acc / data_len
def train():
print("Configuring TensorBoard ")
# 配置 Tensorboard，重新训练时，请将tensorboard⽂件夹删除，不然图会覆盖
tensorboard_dir = 'tensorboard/textcnn'
if not ists(tensorboard_dir):
os.makedirs(tensorboard_dir)
tf.summary.scalar("loss", model.loss)
tf.summary.scalar("accuracy", model.acc)
merged_summary = _all()
writer = tf.summary.FileWriter(tensorboard_dir)
# 配置 Saver
saver = tf.train.Saver()
if not ists(save_dir):
os.makedirs(save_dir)
print("Loading training and ")
# 载⼊训练集与验证集
start_time = time.time()
x_train, y_train = process_file(train_dir, word_to_id, cat_to_id, config.seq_length)
x_val, y_val = process_file(val_dir, word_to_id, cat_to_id, config.seq_length)
time_dif = get_time_dif(start_time)
print("Time usage:", time_dif)
# 创建session
session = tf.Session()
session.run(tf.global_variables_initializer())
writer.add_aph)
print('Training ')
start_time = time.time()
total_batch = 0 # 总批次
加工助剂acr
best_acc_val = 0.0 # 最佳验证集准确率
last_improved = 0 # 记录上⼀次提升批次
require_improvement = 1000 # 如果超过1000轮未提升，提前结束训练
flag = False
for epoch in range(config.num_epochs):
print('Epoch:', epoch + 1)
batch_train = batch_iter(x_train, y_train, config.batch_size)
for x_batch, y_batch in batch_train:
feed_dict = feed_data(x_batch, y_batch, config.dropout_keep_prob)
if total_batch % config.save_per_batch == 0:
# 每多少轮次将训练结果写⼊tensorboard scalar
s = session.run(merged_summary, feed_dict=feed_dict)
writer.add_summary(s, total_batch)
if total_batch % config.print_per_batch == 0:
# 每多少轮次输出在训练集和验证集上的性能
feed_dict[model.keep_prob] = 1.0
loss_train, acc_train = session.run([model.loss, model.acc], feed_dict=feed_dict)
loss_val, acc_val = evaluate(session, x_val, y_val) # todo
if acc_val > best_acc_val:
# 保存最好结果
best_acc_val = acc_val
last_improved = total_batch
saver.save(sess=session, save_path=save_path)
improved_str = '*'
else:
improved_str = ''
time_dif = get_time_dif(start_time)
msg = 'Iter: {0:>6}, Train Loss: {1:>6.2}, Train Acc: {2:>7.2%},' \
+ ' Val Loss: {3:>6.2}, Val Acc: {4:>7.2%}, Time: {5} {6}'
print(msg.format(total_batch, loss_train, acc_train, loss_val, acc_val, time_dif, improved_str)) session.run(model.optim, feed_dict=feed_dict) # 运⾏优化
total_batch += 1
if total_batch - last_improved > require_improvement:
# 验证集正确率长期不提升，提前结束训练
print("No optimization for a long time, ")
flag = True
break# 跳出循环
if flag: # 同上
break
def test():
print("Loading ")
start_time = time.time()
x_test, y_test = process_file(test_dir, word_to_id, cat_to_id, config.seq_length)
session = tf.Session()
session.run(tf.global_variables_initializer())
saver = tf.train.Saver()
print('')
loss_test, acc_test = evaluate(session, x_test, y_test)
msg = 'Test Loss: {0:>6.2}, Test Acc: {1:>7.2%}'
print(msg.format(loss_test, acc_test))
batch_size = 128
data_len = len(x_test)
num_batch = int((data_len - 1) / batch_size) + 1
y_test_cls = np.argmax(y_test, 1)
y_pred_cls = np.zeros(shape=len(x_test), dtype=np.int32) # 保存预测结果
for i in range(num_batch): # 逐批次处理
start_id = i * batch_size
end_id = min((i + 1) * batch_size, data_len)
feed_dict = {
model.input_x: x_test[start_id:end_id],
model.keep_prob: 1.0
}
y_pred_cls[start_id:end_id] = session.run(model.y_pred_cls, feed_dict=feed_dict)
# 评估直流电机编码器
print("Precision, Recall ")
print(metrics.classification_report(y_test_cls, y_pred_cls, target_names=categories))
# 混淆矩阵
print("")
cm = fusion_matrix(y_test_cls, y_pred_cls)
print(cm)
time_dif = get_time_dif(start_time)
print("Time usage:", time_dif)
if__name__ == '__main__':
print('Configuring ')
config = TCNNConfig()
if not ists(vocab_dir): # 如果不存在词汇表，重建
build_vocab(train_dir, vocab_dir, config.vocab_size)
categories, cat_to_id = read_category()
words, word_to_id = read_vocab(vocab_dir)
config.vocab_size = len(words)
model = TextCNN(config)
# train()
test()
# coding: utf-8
from__future__import print_function
import os
import tensorflow as tf
ib.keras as kr
import time
from run_cnn import get_time_dif
from cnn_model import TCNNConfig, TextCNN
ws_loader import read_category, read_vocab
base_dir = 'data/'
vocab_dir = os.path.join(base_dir, '')
save_dir = 'checkpoints/textcnn'
save_path = os.path.join(save_dir, 'best_validation') # 最佳验证结果保存路径
class CnnModel:局部镀锡
def__init__(self):
self.categories, self.cat_to_id = read_category()
self.words, self.word_to_id = read_vocab(vocab_dir)
self.session = tf.Session()
self.session.run(tf.global_variables_initializer())
saver = tf.train.Saver()
def predict(self, message):
# ⽀持不论在python2还是python3下训练的模型都可以在2或者3的环境下运⾏
content = message
data = [self.word_to_id[x] for x in content if x in self.word_to_id]
feed_dict = {
}
y_pred_cls = self.session.del.y_pred_cls, feed_dict=feed_dict)
return self.categories[y_pred_cls[0]]
if__name__ == '__main__':
starttime = time.time()
cnn_model = CnnModel()
test_demo = [' 16-12-08 今年前11个⽉我国进出⼝总值21.83万亿元 ',
'16-12-08 英国知识产权局局长⼀⾏访问黄埔海关（图', '16-12-08 厦门海关启动“互联⽹+⾃主报关”改⾰ ',
'16-12-08 江门海关“宪法⽇” 普法到⼀线（图）',
'16-12-08 27.5公⽄“萌萌哒”果实种⼦闯关被截获（图）',
'⼴州海关推动“主动披露” 体现执法“宽严相济”',
'16-12-07 胡伟在湛江出席全国沿海沿边地区基层反⾛私综合治理现场会（图）',
'16-12-07 锐意改⾰⾼效服务海关⼒助湛江书写蓝⾊经济梦想',
'16-12-07 红其拉甫海关查获4.8千克（图）'] for i in test_demo:
print(cnn_model.predict(i))
print(get_time_dif(starttime))
