久久精品在线视频,99久久一香蕉国产线看观看,青青青爽视频在线观看入口

import numpy as np

start_date = '2024-01-01'
end_date = '2024-12-31'
dates = pd.date_range(start=start_date, end=end_date)

data = {
'Feature1': np.random.randn(len(dates)),
'Feature2': np.random.rand(len(dates)) * 100, # 生成0到100之間的隨機(jī)數(shù)
'Feature3': np.random.randint(1, 100, size=len(dates)) # 生成1到100之間的隨機(jī)整數(shù)
}

df = pd.DataFrame(data, index=dates)
df

在這里我們生成一個(gè)時(shí)序數(shù)據(jù)，接下將使用不同的窗口形式對(duì)該數(shù)據(jù)進(jìn)行時(shí)間窗口劃分。

2. 單輸入單輸出單步預(yù)測(cè)

def prepare_data(data, win_size):

    X = []  

    y = []  



    for i in range(len(data) - win_size):

        temp_x = data[i:i + win_size] 

        temp_y = data[i + win_size]    

        X.append(temp_x)

        y.append(temp_y)



    X = np.asarray(X)

    y = np.asarray(y)

    X = np.expand_dims(X, axis=-1)

    return X, y



win_size = 12

X, y = prepare_data(df['Feature1'].values, win_size)



train_size = int(len(X) * 0.7)  



X_train, X_test = X[:train_size], X[train_size:]

y_train, y_test = y[:train_size], y[train_size:]



print("訓(xùn)練集形狀:", X_train.shape, y_train.shape)

print("測(cè)試集形狀:", X_test.shape, y_test.shape)

單輸入單輸出單步預(yù)測(cè)指的是在時(shí)間序列預(yù)測(cè)中，使用單個(gè)特征（或變量）的歷史觀測(cè)數(shù)據(jù)來(lái)預(yù)測(cè)未來(lái)該輸入特征單個(gè)時(shí)間點(diǎn)的結(jié)果。

在這里數(shù)據(jù)形狀維度的含義為：訓(xùn)練集包含 247 個(gè)樣本，每個(gè)樣本時(shí)間窗口為12，每個(gè)時(shí)間步長(zhǎng)有 1 個(gè)特征；測(cè)試集包含 107 個(gè)樣本，每個(gè)樣本時(shí)間窗口為12，每個(gè)時(shí)間步長(zhǎng)有 1 個(gè)特征。

3.?單輸入單輸出多步預(yù)測(cè)

def prepare_data(data, win_size, steps):

    X = []  

    y = []  



    for i in range(len(data) - win_size - steps + 1):

        temp_x = data[i:i + win_size] 

        temp_y = data[i + win_size:i + win_size + steps]    

        X.append(temp_x)

        y.append(temp_y)



    X = np.asarray(X)

    y = np.asarray(y)

    X = np.expand_dims(X, axis=-1)

    return X, y



win_size = 12

steps = 3  # 修改為多步預(yù)測(cè)，預(yù)測(cè)未來(lái)3個(gè)時(shí)間步

X, y = prepare_data(df['Feature1'].values, win_size, steps)



train_size = int(len(X) * 0.7)  



X_train, X_test = X[:train_size], X[train_size:]

y_train, y_test = y[:train_size], y[train_size:]



print("訓(xùn)練集形狀:", X_train.shape, y_train.shape)

print("測(cè)試集形狀:", X_test.shape, y_test.shape)

單輸入單輸出多步預(yù)測(cè)指的是使用單個(gè)特征的歷史觀測(cè)數(shù)據(jù)來(lái)預(yù)測(cè)未來(lái)連續(xù)多個(gè)時(shí)間點(diǎn)的結(jié)果。

在這里數(shù)據(jù)形狀維度的含義為：訓(xùn)練集包含 246 個(gè)樣本，每個(gè)樣本有 12 個(gè)時(shí)間步長(zhǎng)，每個(gè)時(shí)間步長(zhǎng)有 1 個(gè)特征，同時(shí)輸出該輸入特征未來(lái)三個(gè)時(shí)間點(diǎn)的預(yù)測(cè)結(jié)果；測(cè)試集包含 106 個(gè)樣本，每個(gè)樣本有 12 個(gè)時(shí)間步長(zhǎng)，每個(gè)時(shí)間步長(zhǎng)有 1 個(gè)特征，同時(shí)輸出該輸入特征未來(lái)三個(gè)時(shí)間點(diǎn)的預(yù)測(cè)結(jié)果。

4. 多輸入單輸出單步預(yù)測(cè)

ef prepare_data(data, win_size, target_feature_idx):

    num_features = data.shape[1]

    X = []  

    y = []  

    for i in range(len(data) - win_size):

        temp_x = data[i:i + win_size, :]  

        temp_y = data[i + win_size, target_feature_idx]  

        X.append(temp_x)

        y.append(temp_y)

    X = np.asarray(X)

    y = np.asarray(y)



    return X, y



win_size = 12

target_feature_idx = 0 # 指定待預(yù)測(cè)特征列

X, y = prepare_data(df.values, win_size, target_feature_idx)



train_size = int(len(X) * 0.7)  



X_train, X_test = X[:train_size], X[train_size:]

y_train, y_test = y[:train_size], y[train_size:]



print("訓(xùn)練集形狀:", X_train.shape, y_train.shape)

print("測(cè)試集形狀:", X_test.shape, y_test.shape)

多輸入單輸出單步預(yù)測(cè)是指使用多個(gè)特征的歷史觀測(cè)數(shù)據(jù)來(lái)預(yù)測(cè)未來(lái)單個(gè)時(shí)間點(diǎn)的結(jié)果。

在這里數(shù)據(jù)形狀維度的含義為：訓(xùn)練集包含 247 個(gè)樣本，每個(gè)樣本有 12 個(gè)時(shí)間步長(zhǎng)，每個(gè)時(shí)間步長(zhǎng)有 3 個(gè)特征，同時(shí)有 1 個(gè)目標(biāo)變量；測(cè)試集包含 107 個(gè)樣本，每個(gè)樣本有 12 個(gè)時(shí)間步長(zhǎng)，每個(gè)時(shí)間步長(zhǎng)有 3 個(gè)特征，同時(shí)有 1 個(gè)目標(biāo)變量。

5. 多輸入單輸出多步預(yù)測(cè)

def prepare_data(data, win_size, target_feature_idx, steps):

    num_features = data.shape[1]

    X = []  

    y = []  

    for i in range(len(data) - win_size - steps + 1):

        temp_x = data[i:i + win_size, :]  

        temp_y = data[i + win_size:i + win_size + steps, target_feature_idx]  

        X.append(temp_x)

        y.append(temp_y)

    X = np.asarray(X)

    y = np.asarray(y)



    return X, y



win_size = 12

target_feature_idx = 0 # 指定待預(yù)測(cè)特征列

steps = 3  # 預(yù)測(cè)未來(lái)3個(gè)時(shí)間步

X, y = prepare_data(df.values, win_size, target_feature_idx, steps)



train_size = int(len(X) * 0.7)  



X_train, X_test = X[:train_size], X[train_size:]

y_train, y_test = y[:train_size], y[train_size:]



print("訓(xùn)練集形狀:", X_train.shape, y_train.shape)

print("測(cè)試集形狀:", X_test.shape, y_test.shape)

多輸入單輸出多步預(yù)測(cè)是指使用多個(gè)特征的歷史觀測(cè)數(shù)據(jù)來(lái)預(yù)測(cè)未來(lái)連續(xù)多個(gè)時(shí)間點(diǎn)的結(jié)果。

在這里數(shù)據(jù)形狀維度的含義為：訓(xùn)練集包含 246 個(gè)樣本，每個(gè)樣本有 12 個(gè)時(shí)間步長(zhǎng)，每個(gè)時(shí)間步長(zhǎng)有 3 個(gè)特征，同時(shí)輸出指定預(yù)測(cè)特征序列未來(lái)三個(gè)時(shí)間點(diǎn)的預(yù)測(cè)結(jié)果；測(cè)試集包含 106 個(gè)樣本，每個(gè)樣本有 12 個(gè)時(shí)間步長(zhǎng)，每個(gè)時(shí)間步長(zhǎng)有 3 個(gè)特征，同時(shí)輸出指定預(yù)測(cè)特征序列未來(lái)三個(gè)時(shí)間點(diǎn)的預(yù)測(cè)結(jié)果。

6. 多輸入多輸出單步預(yù)測(cè)

def prepare_data(data, win_size, target_feature_idx):

    num_features = data.shape[1]

    X = []  

    y = []  

    for i in range(len(data) - win_size):

        temp_x = data[i:i + win_size, :]  

        temp_y = data[i + win_size, target_feature_idx]  

        X.append(temp_x)

        y.append(temp_y)

    X = np.asarray(X)

    y = np.asarray(y)



    return X, y



win_size = 12

target_feature_idx = [0,1,2] # 指定待預(yù)測(cè)特征列

X, y = prepare_data(df.values, win_size, target_feature_idx)



train_size = int(len(X) * 0.7)  



X_train, X_test = X[:train_size], X[train_size:]

y_train, y_test = y[:train_size], y[train_size:]



print("訓(xùn)練集形狀:", X_train.shape, y_train.shape)

print("測(cè)試集形狀:", X_test.shape, y_test.shape)

多輸入多輸出單步預(yù)測(cè)是指使用多個(gè)特征的歷史觀測(cè)數(shù)據(jù)來(lái)預(yù)測(cè)未來(lái)多個(gè)特征單個(gè)時(shí)間點(diǎn)的輸出特征結(jié)果。

在這里數(shù)據(jù)形狀維度的含義為：訓(xùn)練集包含 247 個(gè)樣本，每個(gè)樣本有 12 個(gè)時(shí)間步長(zhǎng)，每個(gè)時(shí)間步長(zhǎng)有 3 個(gè)特征，同時(shí)有 3 個(gè)目標(biāo)變量（也就是指定的待預(yù)測(cè)序列）；測(cè)試集包含 107 個(gè)樣本，每個(gè)樣本有 12 個(gè)時(shí)間步長(zhǎng)，每個(gè)時(shí)間步長(zhǎng)有 3 個(gè)特征，同時(shí)有 3 個(gè)目標(biāo)變量（也就是指定的待預(yù)測(cè)序列）。

7. 多輸入多輸出多步預(yù)測(cè)

def prepare_data(data, win_size, target_feature_idx, steps):

    num_features = data.shape[1]

    X = []  

    y = []  

    for i in range(len(data) - win_size - steps + 1):

        temp_x = data[i:i + win_size, :]  

        temp_y = data[i + win_size:i + win_size + steps, target_feature_idx]  

        X.append(temp_x)

        y.append(temp_y)

    X = np.asarray(X)

    y = np.asarray(y)



    return X, y



win_size = 12

target_feature_idx = [0, 1, 2] # 指定待預(yù)測(cè)特征列

steps = 4  # 預(yù)測(cè)未來(lái)4個(gè)時(shí)間步

X, y = prepare_data(df.values, win_size, target_feature_idx, steps)



train_size = int(len(X) * 0.7)  



X_train, X_test = X[:train_size], X[train_size:]

y_train, y_test = y[:train_size], y[train_size:]



print("訓(xùn)練集形狀:", X_train.shape, y_train.shape)

print("測(cè)試集形狀:", X_test.shape, y_test.shape)

多輸入多輸出多步預(yù)測(cè)是指使用多個(gè)特征的歷史觀測(cè)數(shù)據(jù)來(lái)預(yù)測(cè)未來(lái)多個(gè)特征多個(gè)時(shí)間點(diǎn)的輸出特征結(jié)果。

在這里數(shù)據(jù)形狀維度的含義為：訓(xùn)練集包含 245 個(gè)樣本，每個(gè)樣本有 12 個(gè)時(shí)間步長(zhǎng)，每個(gè)時(shí)間步長(zhǎng)有 3 個(gè)特征，同時(shí)有 3 個(gè)目標(biāo)變量（也就是指定的待預(yù)測(cè)序列），同時(shí)輸出每個(gè)目標(biāo)變量未來(lái)三個(gè)時(shí)間點(diǎn)的預(yù)測(cè)結(jié)果；測(cè)試集包含 106 個(gè)樣本，每個(gè)樣本有 12 個(gè)時(shí)間步長(zhǎng)，每個(gè)時(shí)間步長(zhǎng)有 3 個(gè)特征，同時(shí)有 3 個(gè)目標(biāo)變量（也就是指定的待預(yù)測(cè)序列），同時(shí)輸出每個(gè)目標(biāo)變量未來(lái)三個(gè)時(shí)間點(diǎn)的預(yù)測(cè)結(jié)果。

文章轉(zhuǎn)自微信公眾號(hào)@Python機(jī)器學(xué)習(xí)AI