Question

为了将训练数据分为输入（x）和输出（y），我必须执行以下命令：

X, y = train[:, 0: -1], train[:, -1]

很遗憾，它不起作用，并且在打印train[:, 0: -1]时会产生以下错误：

> TypeError: '(slice(None, None, None), slice(0, -1, None))' is an invalid key

我尝试用以下命令替换此命令：

X, y = train[:][0: -1], train[:][-1]

但是我很确定它不会提供相同的输出，因为（在我的情况下）不合逻辑的输入和一个输出都是

以下是带有代码示例的最小可复制代码：

from pandas import DataFrame
from pandas import datetime
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
import numpy

O = [0.701733664614, 0.699495411782, 0.572129320819, 0.613315597684, 0.58079660603, 0.596638918579, 0.48453382119]
Ab = [datetime(2018, 12, 11, 14, 0), datetime(2018, 12, 21, 10, 0), datetime(2018, 12, 21, 14, 0), datetime(2019, 1, 1, 10, 0), datetime(2019, 1, 1, 14, 0), datetime(2019, 1, 11, 10, 0), datetime(2019, 1, 11, 14, 0)]

data = DataFrame(numpy.column_stack([O, Ab]),
                 columns=['ndvi', 'datetime'])

def fit_lstm(train, batch_size, nb_epoch, neurons):
    X, y = train[:, 0: -1], train[:, -1]
    X = X.values.reshape(X.shape[0], 1, X.shape[1])
    model = Sequential()
    model.add(LSTM(neurons, batch_input_shape=(batch_size, X.shape[1], X.shape[2]), stateful=True))
    model.add(Dense(1))
    model.compile(loss='mean_squared_error', optimizer='adam')
    for i in range(nb_epoch):
        model.fit(X, y, epochs=1, batch_size=batch_size, verbose=0, shuffle=False)
        model.reset_states()
    return model

train, test = data.values[0:-2], data.values[-2:]

print (train[:, 0:-1])

我要解决的问题是适合LSTM模型：

lstm_model = fit_lstm(train, 1, 3000, 4)

也许，在这种情况下，我必须使用shift()，以通过最后一个时间步作为输入，而将当前时间步作为输出？像这样：

shift_steps = 1
train_targets = train.shift(-shift_steps)
X, y = train, train_targets

Answer 1

您正在使用DataFrame，并且切片有所不同。

假设“ nvdi”列具有功能，而datetime列则是每种训练数据的预期结果，则需要将它们称为：

X = train['nvdi']
y = train['datetime']

我在数据中只看到两列。

Answer 2

我认为我发现了我的错误：语法与我的数据集不兼容。

正如我在问题的最后编辑中提到的那样，我使用了shift。它是这样工作的：


def fit_lstm(train, batch_size, nb_epoch, neurons):
    X, y = train, train_targets
    X = X.values.reshape(X.shape[0], 1, X.shape[1])
    model = Sequential()
    model.add(LSTM(neurons, batch_input_shape=(batch_size, X.shape[1], X.shape[2]), stateful=True))
    model.add(Dense(1))
    model.compile(loss='mean_squared_error', optimizer='adam')
    for i in range(nb_epoch):
        model.fit(X, y, epochs=1, batch_size=batch_size, verbose=0, shuffle=False)
        model.reset_states()
    return model

train, test = data.values[0:-2], data.values[-2:]
shift_steps = 1
train_targets = data.shift(-shift_steps).values[0:-2]

||

/

print (train)
[0.70173366 0.69949541 0.57212932 0.6133156  0.58079661]

print (train_targets)
[0.69949541 0.57212932 0.6133156  0.58079661 0.59663892]

非常感谢@ 00和@Sebas的帮助：）

是array [:, 0：-1] = array [：] [0：-1]？

2 个答案: