深度学习手写简单的lstm

代码如下；

import pandas as pd
import numpy as np
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten, LSTM, TimeDistributed, RepeatVector
from keras.layers.normalization.batch_normalization_v1 import BatchNormalization
from keras.optimizers import Adam
from keras.callbacks import EarlyStopping, ModelCheckpoint
from sklearn.preprocessing import StandardScaler, MinMaxScaler
import matplotlib.pyplot as plt
n_in = 168  # 历史数量
n_out = 24  # 预测数量
n_features = 1
# n_test = 1
n_val = 1
n_epochs = 250


# 导入数据
def load_stw_data() -> pd.DataFrame:   #在def那一行后面会加一个->。这个玩意儿有个专门的名词叫 type hint， 即类型提示
    df_stw = pd.read_excel('data3.xlsx')
    df_stw.columns = ['BillingDate', 'VolumnHL']

    return df_stw


# MinMaxScaler数据归一化，可以帮助网络模型更快的拟合，稍微有一些提高准确率的效果
def minmaxscaler(data: pd.DataFrame) -> pd.DataFrame:
    volume = data.VolumnHL.values
    volume = volume.reshape(len(volume), 1)
    volume = scaler.fit_transform(volume)
    volume = volume.reshape(len(volume), )
    data['VolumnHL'] = volume

    return data


# 划分训练数据集和验证数据集,这里需要注意的是我么需要预测的数据是不可以出现在训练中的，切记。
def split_data(x, y, n_test: int):
    x_train = x[:-n_val - n_out + 1]
    x_val = x[-n_val:]
    y_train = y[:-n_val - n_out + 1]
    y_val = y[-n_val:]

    return x_train, y_train, x_val, y_val


# 划分X和Y
def build_train(train, n_in, n_out):
    train = train.drop(["BillingDate"], axis=1)
    X_train, Y_train = [], []
    for i in range(train.shape[0] - n_in - n_out + 1):
        X_train.append(np.array(train.iloc[i:i + n_in]))
        Y_train.append(np.array(train.iloc[i + n_in:i + n_in + n_out]["VolumnHL"]))

    return np.array(X_train), np.array(Y_train)


# 构建最简单的LSTM
def build_lstm(n_in: int, n_features: int):
    model = Sequential()
    model.add(LSTM(12, activation='relu', input_shape=(n_in, n_features)))
    model.add(Dropout(0.3))
    model.add(Dense(n_out))
    model.compile(optimizer='adam', loss='mae')

    return model


# 模型拟合
def model_fit(x_train, y_train, x_val, y_val, n_features):
    model = build_lstm(
        n_in=n_in,
        n_features=1
    )
    model.compile(loss='mae', optimizer='adam')
    model.fit(x_train, y_train, epochs=n_epochs, batch_size=128, verbose=1, validation_data=(x_val, y_val))
    m = model.evaluate(x_val, y_val)
    print(m)

    return model


data = load_stw_data()
scaler = MinMaxScaler(feature_range=(0, 1))
data = minmaxscaler(data)
data_copy = data.copy()
x, y = build_train(data_copy, n_in, n_out)
x_train, y_train, x_val, y_val = split_data(x, y, n_val)
model = build_lstm(n_in, 1)
model = model_fit(x_train, y_train, x_val, y_val, 1)
predict = model.predict(x_val)
# predict = model.predict(x_val)
validation = scaler.inverse_transform(predict)[0]
validation
actual = scaler.inverse_transform(y_val)[0]
actual
predict = validation
actual = actual
x = [x for x in range(24)]
fig, ax = plt.subplots(figsize=(15,5),dpi = 300)
ax.plot(x, predict, linewidth=2.0,label = "predict")
ax.plot(x, actual, linewidth=2.0,label = "actual")
ax.legend(loc=2);
# ax.set_title(bf_name)
plt.ylim((0, 900000))
plt.grid(linestyle='-.')
plt.show()