【发布时间】：2020-04-08 11:38:35
【问题描述】：
第一个对不起我的英语不好。我正在创建一个用于预测股票价格的 Keras LSTM 模型。这是我的代码：
from tqdm import tqdm
import numpy as np
import pandas as pd
from keras.models import Sequential
from keras.layers.recurrent import LSTM
from keras.layers.core import Dense,Activation,Dropout,Flatten,Reshape
from sklearn.preprocessing import MinMaxScaler
import keras as kr
from sklearn.model_selection import train_test_split
from keras import optimizers
veri =pd.read_csv("eurusd.csv")
veri['trh'] = pd.to_datetime(veri.trh, format='%d.%m.%Y')
########################
del veri['puan']
del veri['yuzde']
del veri['sira']
del veri['trh']
df_train, df_test = train_test_split(veri, train_size=0.8, test_size=0.2, shuffle=False)
print("Train and Test size", len(df_train), len(df_test))
x = df_train.loc[:,:].values
scaler = MinMaxScaler(feature_range=(0,1))
x_train = scaler.fit_transform(x)
x_test = scaler.transform(df_test.loc[:,:])
TIME_STEPS=7
BATCH_SIZE=128
def build_timeseries(mat, y_col_index):
# y_col_index tahmin etmek istediğimiz değerin sütun numarası
# total number of time-series samples would be len(mat) - TIME_STEPS
dim_0 = mat.shape[0] - TIME_STEPS #1328-7 gibi bir şey
dim_1 = mat.shape[1]
x = np.zeros((dim_0, TIME_STEPS, dim_1))
y = np.zeros((dim_0,))
for i in tqdm(range(dim_0)):
x[i] = mat[i:TIME_STEPS + i]
y[i] = mat[TIME_STEPS + i, y_col_index]
print("length of time-series i/o", x.shape, y.shape)
return x, y
def trim_dataset(mat, batch_size):
"""
trims dataset to a size that's divisible by BATCH_SIZE
"""
no_of_rows_drop = mat.shape[0]%batch_size
if(no_of_rows_drop > 0):
return mat[:-no_of_rows_drop]
else:
return mat
x_t, y_t = build_timeseries(x_train, 0)
#x_t =3 boyutlu besleme verileri
#y_t =de sonuç satırının timestepsten sonraki kısmı(1. değişkeni aldık)
x_t = trim_dataset(x_t, BATCH_SIZE)#xtrain
y_t = trim_dataset(y_t, BATCH_SIZE)#ytrain(sonuc)
x_temp, y_temp = build_timeseries(x_test, 0)
x_val, x_test_t = np.split(trim_dataset(x_temp, BATCH_SIZE),2)
y_val, y_test_t = np.split(trim_dataset(y_temp, BATCH_SIZE),2)
model = Sequential()
model.add(LSTM(100, batch_input_shape=(BATCH_SIZE, TIME_STEPS, x_t.shape[2]), dropout=0.0, recurrent_dropout=0.0, stateful=True, kernel_initializer='random_uniform'))
model.add(Dropout(0.2))
model.add(Dense(20, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='mean_squared_error', optimizer=kr.optimizers.rmsprop(0.01))
csv_logger = kr.callbacks.CSVLogger('sonuclar.log')
history = model.fit(x_t, #train girdiler
y_t, #train çıktılar
epochs=200,
verbose=2,
batch_size=BATCH_SIZE,
shuffle=False,
validation_data=((trim_dataset(x_val, BATCH_SIZE)),
(trim_dataset(y_val, BATCH_SIZE))),
callbacks=[csv_logger])
grafik1=model.predict(x_test_t)
这就是我的结果：
.
.
.
.
Epoch 192/200
- 0s - loss: 3.8528e-04 - val_loss: 1.0298e-04
Epoch 193/200
- 0s - loss: 3.1330e-04 - val_loss: 3.7064e-04
Epoch 194/200
- 0s - loss: 9.6561e-04 - val_loss: 1.4455e-04
Epoch 195/200
- 0s - loss: 2.7916e-04 - val_loss: 3.4224e-04
Epoch 196/200
- 0s - loss: 8.4071e-04 - val_loss: 4.0075e-04
Epoch 197/200
- 0s - loss: 4.9036e-04 - val_loss: 1.1518e-04
Epoch 198/200
- 0s - loss: 3.2914e-04 - val_loss: 1.2514e-04
Epoch 199/200
- 0s - loss: 0.0010 - val_loss: 3.6713e-04
Epoch 200/200
- 0s - loss: 4.0553e-04 - val_loss: 1.7644e-04
2020-04-08 14:23:10.216020: W tensorflow/core/common_runtime/base_collective_executor.cc:217] BaseCollectiveExecutor::StartAbort Invalid argument: Specified a list with shape [128,4] from a tensor with shape [32,4]
[[{{node lstm_1/TensorArrayUnstack/TensorListFromTensor}}]]
Traceback (most recent call last):
File "/home/phylo/PycharmProjects/Keras1/kerassecond.py", line 80, in <module>
grafik1=model.predict(x_test_t)
File "/home/phylo/.local/lib/python3.6/site-packages/keras/engine/training.py", line 1462, in predict
callbacks=callbacks)
File "/home/phylo/.local/lib/python3.6/site-packages/keras/engine/training_arrays.py", line 324, in predict_loop
batch_outs = f(ins_batch)
File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/keras/backend.py", line 3727, in __call__
outputs = self._graph_fn(*converted_inputs)
File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/eager/function.py", line 1551, in __call__
return self._call_impl(args, kwargs)
File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/eager/function.py", line 1591, in _call_impl
return self._call_flat(args, self.captured_inputs, cancellation_manager)
File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/eager/function.py", line 1692, in _call_flat
ctx, args, cancellation_manager=cancellation_manager))
File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/eager/function.py", line 545, in call
ctx=ctx)
File "/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/eager/execute.py", line 67, in quick_execute
six.raise_from(core._status_to_exception(e.code, message), None)
File "<string>", line 3, in raise_from
tensorflow.python.framework.errors_impl.InvalidArgumentError: Specified a list with shape [128,4] from a tensor with shape [32,4]
[[node lstm_1/TensorArrayUnstack/TensorListFromTensor (defined at home/phylo/.local/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3009) ]] [Op:__inference_keras_scratch_graph_10579]
Function call stack:
keras_scratch_graph
Process finished with exit code 1
当我将批量大小更改为 128 并且我的预测代码 (grafik1=model.predict(x_test_t)) 时，我收到此错误，但如果我尝试将我的 BATC_SIZE 更改为 32，此代码不会给出任何错误。为什么？我测试了不同的值，我发现，128 Batchsize 是我最好的。我该如何解决这个问题？
【问题讨论】：
标签： python tensorflow keras deep-learning lstm