小白的第一个RNN（情感分析模型）

平台：window10，python3.11.4，pycharm

框架：keras

编写日期：20230903

数据集：英语，自编，训练集和测试集分别有4个样本，标签有积极和消极两种

环境搭建

新建文件夹，进入目录

创建虚拟环境

virtualenv venv

激活虚拟环境

venv\Scripts\activate

安装依赖库

pip install tensorflow

代码编写

目录下创建main.py，进入pycharm打开文件夹，编写代码

包引入

import numpy as np
from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences
from keras.models import Sequential
from keras.layers import SimpleRNN, Dense

数据集处理

# 训练集
train_texts = ['I love this movie',
               'This is the worst film I have ever seen.',
               'An enjoyable and thought-provoking experience.',
               "I think it is boring"
               ]

train_labels = np.array([1, 0, 1, 0]) # 0代表消极，1代表积极

# 测试集
test_texts = ["What a waste of my time",
              "One of the best movies I've seen in a long time",
              "Amazing acting!",
              "This movie look awful"
              ]
test_labels = np.array([0, 1, 1, 0])

# 构建分词器
tokenizer = Tokenizer(num_words=100)
# 用训练集与测试集训练分词器
tokenizer.fit_on_texts(train_texts + test_texts)
# 数据集序列化，将文本转成数字，便于机器处理
train_sequences = tokenizer.texts_to_sequences(train_texts)
test_sequences = tokenizer.texts_to_sequences(test_texts)
# 数据填充到20，超过的就截断，post：在末尾填充
# 由于每个训练文本有不同的单词数，需要统一
train_data = pad_sequences(train_sequences, maxlen=20, padding='post')
test_data = pad_sequences(test_sequences, maxlen=20, padding='post')

模型搭建和训练

# 创建一个线性模型容器
model = Sequential()
#添加RNN层，神经元数量为100，输入数据形状为（20，1）
model.add(SimpleRNN(100, input_shape=(20, 1)))
# 添加1个输出，激活函数为sigmoid的全连接层
model.add(Dense(1, activation='sigmoid'))
# 优化器：Adam，损失计算方法：二元交叉熵，评估依据：准确率
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
# 输出模型结构
model.summary()
#训练模型，训练5轮，每次训练2个样本
model.fit(train_data, train_labels, epochs=5, batch_size=2, validation_data=(test_data, test_labels))

模型评估

# 打印评估信息
print('Evaluating the model...')
#进行评估
model.evaluate(test_data, test_labels)

所有代码集合

import numpy as np
from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences
from keras.models import Sequential
from keras.layers import SimpleRNN, Dense

train_texts = ['I love this movie',
               'This is the worst film I have ever seen.',
               'An enjoyable and thought-provoking experience.',
               "I think it is boring"
               ]
train_labels = np.array([1, 0, 1, 0])

test_texts = ["What a waste of my time",
              "One of the best movies I've seen in a long time",
              "Amazing acting!",
              "This movie look awful"
              ]
test_labels = np.array([0, 1, 1, 0])

tokenizer = Tokenizer(num_words=1000)
tokenizer.fit_on_texts(train_texts + test_texts)

train_sequences = tokenizer.texts_to_sequences(train_texts)
test_sequences = tokenizer.texts_to_sequences(test_texts)

train_data = pad_sequences(train_sequences, maxlen=20, padding='post')
test_data = pad_sequences(test_sequences, maxlen=20, padding='post')

model = Sequential()
model.add(SimpleRNN(100, input_shape=(20, 1)))
model.add(Dense(1, activation='sigmoid'))
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
model.summary()

model.fit(train_data, train_labels, epochs=5, batch_size=2, validation_data=(test_data, test_labels))

print('Evaluating the model...')
model.evaluate(test_data, test_labels)

运行图片截取

文件目录

控制台