lstm单步预测

# lstm单步预测
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
import tensorflow as tf
from sklearn.preprocessing import StandardScaler
from tensorflow.keras import layers, Input, optimizers

# 导入数据
data = pd.read_csv(r'D:\me\自己学习\LSTM\stock.csv')
df = pd.DataFrame(data, columns=['date', 'close'])
# print(df.head)

# 创建数据集
dataset = df["close"].values


def data_set(dataset, lookback):  # 创建时间序列数据样本
    dataX, dataY = [], []
    for i in range(len(dataset) - lookback - 1):
        a = dataset[i:(i + lookback)]
        dataX.append(a)
        dataY.append(dataset[i + lookback])
    return np.array(dataX), np.array(dataY)


# 数据标准化
st = StandardScaler()
dataset_st = st.fit_transform(dataset.reshape(-1, 1))
# print("标准化", dataset_st)

# # 划分训练集和测试集
train_size = int(len(dataset_st) * 0.7)
test_size = len(dataset_st) - train_size
train, test = dataset_st[0:train_size], dataset_st[train_size:len(dataset_st)]
# print(len(train))
# print(len(test))

# # 根据划分的训练集测试集生成需要的时间序列样本数据
lookback = 1
trainX, trainY = data_set(train, lookback)
testX, testY = data_set(test, lookback)
# print('trianX:', trainX.shape, trainY.shape)
# print(trainX)

# # 构建lstm模型
input_shape = Input(shape=(trainX.shape[1], trainX.shape[2]))
lstm1 = layers.LSTM(32, return_sequences=1)(input_shape)
print("lstm1:", lstm1.shape)
lstm2 = layers.LSTM(64, return_sequences=0)(lstm1)
print("lstm2:", lstm2.shape)
dense1 = layers.Dense(64, activation="relu")(lstm2)
print("dense:", dense1.shape)
dropout = layers.Dropout(rate=0.2)(dense1)
print("dropout:", dropout.shape)
ouput_shape = layers.Dense(1, activation="relu")(dropout)
lstm_model = tf.keras.Model(input_shape, ouput_shape)
lstm_model.compile(loss="mean_squared_error", optimizer="Adam", metrics=["mse"])
history = lstm_model.fit(trainX, trainY, batch_size=16, epochs=10, validation_split=0.1, verbose=1)

lstm_model.summary()
#
# # 预测测试集
predict_trainY = lstm_model.predict(trainX)
predict_testY = lstm_model.predict(testX)
# # 反标准化
# # trainY=st.inverse_transform(predict_trainY)
testY_real = st.inverse_transform(testY)
testY_predict = st.inverse_transform(predict_testY)
print("Y:", testY_predict, testY_predict.shape)
print("Y222:", testY_real, testY_real.shape)
plt.figure(figsize=(12, 8))
plt.plot(testY_predict, "b", label="predict")
plt.plot(testY_real, "r", label="real")
plt.legend()
plt.show()