【机器学习实战】三、使用LSTM实现IBM股票价格预测

使用LSTM实现IBM股票价格预测

一、导入第三方库

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.preprocessing import MinMaxScaler
from keras.models import Sequential
from keras.layers import Dense, LSTM, Dropout
from keras.optimizers import SGD

二、加载历史数据文件

dataset = pd.read_csv("datasets_8388_11883_IBM_2006-01-01_to_2018-01-01.csv",index_col='Date',parse_dates=['Date'])
dataset.head()

dataset.shape
# 输出：(3020, 6)

# 划分数据集
train_set = dataset[:'2016'].iloc[:,1:2].values # 划分训练集选取0-2016条数据，获取到第一列数据
test_set = dataset['2017':].iloc[:,1:2].values # 划分测试集选取2017-3020条数据 ，获取第一列数据

train_set.shape
# 输出：(2769, 1)

test_set.shape
# 输出：(251, 1)

三、定义显示函数

def plot_predictions(test_result, predict_result):
    """
    test_result:真实值
    predict_result: 预测值
    """
    plt.plot(test_result,color='red',label='IBM True Stock Price')
    plt.plot(predict_result,color='blue',label='IBM Predicted Stock Price')
    plt.title('IBM Stock Price')
    plt.xlabel('Time')
    plt.ylabel('Stock Price')
    plt.legend() # 加上图例
    plt.show()

四、绘制训练集和测试集的数据

dataset['High'][:'2016'].plot(figsize=(16,4),legend=True)
dataset['High']['2017':].plot(figsize=(16,4),legend=True)
plt.title('IBM Stock Price')
plt.legend(['Train set(before 2016)','Test set(after 2017)'])
plt.show()

# 正则化：将每一维德特征映射到指定德区间： 【0，1】
sc = MinMaxScaler(feature_range = [0,1]) # 调用sklearn的归一化方法MinMaxScaler
train_set_scaled = sc.fit_transform(train_set)

# 创建序列数据集（训练和测试）
# 60个时间步为一个样本 ，1个输出
X_train = [] # 训练数据列表
y_train = [] # 训练数据集标签
for i in range(60, 2769) : # 以60个数据集为一个滑动窗口
    X_train.append(train_set_scaled[i-60:i,0])
    y_train.append(train_set_scaled[i,0])
X_train, y_train = np.array(X_train) , np.array(y_train) # numpyl类型变换

X_train.shape
# 输出：(2709, 60)

X_train[0]

#LSTM的输入:(samples,sequence_length,features)
#reshape:训练集(2709,60) ——> (2709,60,1)
X_train = np.reshape(X_train,(X_train.shape[0],X_train.shape[1],1))
X_train.shape
# 输出：(2709, 60, 1)

五、搭建LSTM模型，进行训练和预测

model = Sequential()

# LSTM 第一层
model.add(LSTM(128,return_sequences=True,input_shape=(X_train.shape[1],1)))
model.add(Dropout(0.2))

# LSTM 第二层
model.add(LSTM(128,return_sequences=True))
model.add(Dropout(0.2))

# LSTM第三层
model.add(LSTM(128))
model.add(Dropout(0.2))

# Dense层
model.add(Dense(units=1))

# 模型编译
model.compile(optimizer='rmsprop',loss='mse')
# 模型训练
model.fit(X_train, y_train,epochs=20,batch_size=32)

六、构建数据集，进行训练

dataset_total = pd.concat((dataset['High'][:'2016'],dataset['High']['2017':]),axis = 0)
dataset_total.shape
# 输出：(3020,)

dataset_total

inputs = dataset_total[len(train_set):].values
inputs = inputs.reshape(-1,1)
inputs.shape
# 输出(251, 1)
inputs_scaled = sc.fit_transform(inputs)

七、构建测试集X_test，进行股价预测

dataset_total = pd.concat((dataset['High'][:'2016'],dataset['High']['2017':]),axis=0)
# 获取输入数据
inputs = dataset_total[len(dataset_total) - len(test_set) - 60 :].values

inputs

# 归一化
inputs = inputs.reshape(-1,1)
inputs = sc.transform(inputs)

inputs.shape
# 输出：(311, 1)

# 准备测试集X_test ，进行股价预测
X_test = []
for i in range(60,311):
    X_test.append(inputs[i-60:i,0])
X_test = np.array(X_test) # numpy类型转换
X_test.shape
# 输出(251, 60)

X_test = np.reshape(X_test,(X_test.shape[0],X_test.shape[1],1))

X_test.shape
# 输出：(251, 60, 1)

predict_test = model.predict(X_test) # 预测
predict_test.shape

predict_stock_price = sc.inverse_transform(predict_test)
predict_stock_price

# 绘制测试结果和预测结果
plot_predictions(test_set,predict_stock_price)