正确形状的keras LSTM进料输入

Question

我正在从具有以下形状的熊猫数据框中获取一些数据

df.head()
>>>
Value USD   Drop 7  Up 7    Mean Change 7   Change      Predict
0.06480     2.0     4.0     -0.000429       -0.00420    4
0.06900     1.0     5.0     0.000274        0.00403     2
0.06497     1.0     5.0     0.000229        0.00007     2
0.06490     1.0     5.0     0.000514        0.00200     2
0.06290     2.0     4.0     0.000229        -0.00050    3

前5列旨在作为X并预测y。这就是我预处理模型数据的方式

from keras.models import Sequential
from keras.callbacks import EarlyStopping
from keras.callbacks import ModelCheckpoint
import pandas as pd
from sklearn.model_selection import train_test_split
import numpy as np
from sklearn.metrics import accuracy_score
from keras.layers import LSTM
from sklearn import preprocessing

# Convert a Pandas dataframe to the x,y inputs that TensorFlow needs
def to_xy(df, target):
    result = []
    for x in df.columns:
        if x != target:
            result.append(x)
    # find out the type of the target column.  Is it really this hard? :(
    target_type = df[target].dtypes
    target_type = target_type[0] if hasattr(target_type, '__iter__') else target_type
    # Encode to int for classification, float otherwise. TensorFlow likes 32 bits.
    if target_type in (np.int64, np.int32):
        # Classification
        dummies = pd.get_dummies(df[target])
        return df.as_matrix(result).astype(np.float32), dummies.as_matrix().astype(np.float32)
    else:
        # Regression
        return df.as_matrix(result).astype(np.float32), df.as_matrix([target]).astype(np.float32)

# Encode text values to indexes(i.e. [1],[2],[3] for red,green,blue).
def encode_text_index(df, name):
    le = preprocessing.LabelEncoder()
    df[name] = le.fit_transform(df[name])
    return le.classes_

df['Predict'].value_counts()
>>>
4    1194
3     664
2     623
0     405
1      14
Name: Predict, dtype: int64

predictions = encode_text_index(df, "Predict")
predictions
>>>
array([0, 1, 2, 3, 4], dtype=int64)

X,y = to_xy(df,"Predict")
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20, shuffle=False)

X_train
>>>
array([[ 6.4800002e-02,  2.0000000e+00,  4.0000000e+00, -4.2857142e-04,
        -4.1999999e-03],
       [ 6.8999998e-02,  1.0000000e+00,  5.0000000e+00,  2.7414286e-04,
         4.0300000e-03],
       [ 6.4970002e-02,  1.0000000e+00,  5.0000000e+00,  2.2857143e-04,
         7.0000002e-05],
       ...,
       [ 9.5987000e+02,  5.0000000e+00,  2.0000000e+00, -1.5831429e+01,
        -3.7849998e+01],
       [ 9.9771997e+02,  5.0000000e+00,  2.0000000e+00, -1.6948572e+01,
        -1.8250000e+01],
       [ 1.0159700e+03,  5.0000000e+00,  2.0000000e+00, -1.3252857e+01,
        -7.1700001e+00]], dtype=float32)

y_train
>>>
array([[0., 0., 0., 0., 1.],
       [0., 0., 1., 0., 0.],
       [0., 0., 1., 0., 0.],
       ...,
       [0., 0., 0., 0., 1.],
       [0., 0., 0., 0., 1.],
       [0., 0., 0., 0., 1.]], dtype=float32)

X_train[1]
>>>
array([6.8999998e-02, 1.0000000e+00, 5.0000000e+00, 2.7414286e-04,
       4.0300000e-03], dtype=float32)

X_train.shape
>>>
(2320, 5)

X_train[1].shape
>>>
(5,)

最后是LSTM模型（也可能不是最好的写法，所以如果是这样的话，也将感谢对内层的重写）

model = Sequential()
#model.add(LSTM(128, dropout=0.2, recurrent_dropout=0.2, input_shape=(None, 1)))
model.add(LSTM(50, dropout=0.2, return_sequences=True, input_shape=X_train.shape))
model.add(LSTM(50, dropout=0.2, return_sequences=True))
model.add(LSTM(50, dropout=0.2, return_sequences=True))
model.add(LSTM(50, dropout=0.2, return_sequences=True))
#model.add(Dense(50, activation='relu'))
model.add(Dense(y_train.shape[1], activation='softmax'))

#model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
#model.fit(X_train, y_train, epochs=1000)

model.compile(loss='categorical_crossentropy', optimizer='adam')
monitor = EarlyStopping(monitor='val_loss', min_delta=1e-2, patience=15, verbose=1, mode='auto')
checkpointer = ModelCheckpoint(filepath="best_weights.hdf5", verbose=0, save_best_only=True) # save best model

model.fit(X_train, y_train, validation_data=(X_test, y_test), callbacks=[monitor,checkpointer], verbose=2, epochs=1000)
model.load_weights('best_weights.hdf5') # load weights from best model

运行此将引发此错误

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-67-a17835a382f6> in <module>()
     15 checkpointer = ModelCheckpoint(filepath="best_weights.hdf5", verbose=0, save_best_only=True) # save best model
     16 
---> 17 model.fit(X_train, y_train, validation_data=(X_test, y_test), callbacks=[monitor,checkpointer], verbose=2, epochs=1000)
     18 model.load_weights('best_weights.hdf5') # load weights from best model

c:\users\samuel\appdata\local\programs\python\python35\lib\site-packages\keras\engine\training.py in fit(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, **kwargs)
    948             sample_weight=sample_weight,
    949             class_weight=class_weight,
--> 950             batch_size=batch_size)
    951         # Prepare validation data.
    952         do_validation = False

c:\users\samuel\appdata\local\programs\python\python35\lib\site-packages\keras\engine\training.py in _standardize_user_data(self, x, y, sample_weight, class_weight, check_array_lengths, batch_size)
    747             feed_input_shapes,
    748             check_batch_axis=False,  # Don't enforce the batch size.
--> 749             exception_prefix='input')
    750 
    751         if y is not None:

c:\users\samuel\appdata\local\programs\python\python35\lib\site-packages\keras\engine\training_utils.py in standardize_input_data(data, names, shapes, check_batch_axis, exception_prefix)
    125                         ': expected ' + names[i] + ' to have ' +
    126                         str(len(shape)) + ' dimensions, but got array '
--> 127                         'with shape ' + str(data_shape))
    128                 if not check_batch_axis:
    129                     data_shape = data_shape[1:]

ValueError: Error when checking input: expected lstm_48_input to have 3 dimensions, but got array with shape (2320, 5)

我已经尝试了许多X_train输入形状的变化，但是每个输入形状都会引发一些错误，我还检查了Keras文档，但尚不清楚如何将数据馈送到模型中

尝试1号尝试

首先是重塑X_train

data = np.resize(X_train,(X_train.shape[0],1,X_train.shape[1]))
model.add(LSTM(50, dropout=0.2, return_sequences=True, input_shape=data.shape))

这失败并出现错误

ValueError: Input 0 is incompatible with layer lstm_52: expected ndim=3, found ndim=4 

建议我以

model.add(LSTM(50, dropout=0.2, return_sequences=True, input_shape=X_train.shape[1:]))

引发相同的错误

ValueError: Input 0 is incompatible with layer lstm_63: expected ndim=3, found ndim=2

妊娠2

使用熊猫的默认X，y

y = df['Predict']
X = df[['Value USD', 'Drop 7', 'Up 7', 'Mean Change 7', 'Change']]

X = np.array(X)
y = np.array(y)

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20, shuffle=False)

LSTM还期望以下列方式输入 (batch_size, timesteps, input_dim)

所以我尝试了这个

model.add(LSTM(50, dropout=0.2, return_sequences=True, input_shape=(100, 100, X_train.shape)))

引发此错误

TypeError: Error converting shape to a TensorShape: int() argument must be a string, a bytes-like object or a number, not 'tuple'.

和另一种方式

model.add(LSTM(50, dropout=0.2, return_sequences=True, input_shape=(100, 100, X_train[1].shape)))

返回相同的错误

TypeError: Error converting shape to a TensorShape: int() argument must be a string, a bytes-like object or a number, not 'tuple'.

Answer 1

输入形状假定为格式(no_of_samples,no_of_timesteps,features)

在这里你只有(no_of_samples,features)

您可以在构建网络之前使用 numpy 调整训练数据的大小

data = np.resize(X_train,(X_train.shape[0],1,X_train.shape[1]))

希望这可以帮助