按照我上一篇文章中的建议，我使用 lib KERAS 重写了用于进行时间序列分析的脚本，但在模型中获得了以下输出。
在循环网络中，输入形状应该类似于(批量大小、时间步长、输入特征)。
输出
Traceback (most recent call last):
File "rnrs.py", line 114, in <module>
model = train_model(get_model(), X_train, Y_train, (X_dev, Y_dev), [plot_losses])
File "rnrs.py", line 111, in train_model
model.fit(X_train, Y_train, epochs=200, batch_size=1024, validation_data=validation, callbacks=callbacks, shuffle=False)
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\keras\engine\training.py", line 1213, in fit
self._make_train_function()
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\keras\engine\training.py", line 316, in _make_train_function
loss=self.total_loss)
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\keras\legacy\interfaces.py", line 91, in wrapper
return func(*args, **kwargs)
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\keras\backend\tensorflow_backend.py", line 75, in symbolic_fn_wrapper
return func(*args, **kwargs)
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\keras\optimizers.py", line 543, in get_updates
p_t = p - lr_t * m_t / (K.sqrt(v_t) + self.epsilon)
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow_core\python\ops\math_ops.py", line 903, in binary_op_wrapper
y, dtype_hint=x.dtype.base_dtype, name="y")
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow_core\python\framework\ops.py", line 1242, in convert_to_tensor_v2
as_ref=False)
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow_core\python\framework\ops.py", line 1296, in internal_convert_to_tensor
ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow_core\python\framework\constant_op.py", line 286, in _constant_tensor_conversion_function
return constant(v, dtype=dtype, name=name)
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow_core\python\framework\constant_op.py", line 227, in constant
allow_broadcast=True)
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow_core\python\framework\constant_op.py", line 265, in _constant_impl
allow_broadcast=allow_broadcast))
File "C:\Users\luis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow_core\python\framework\tensor_util.py", line 437, in make_tensor_proto
raise ValueError("None values not supported.")
ValueError: None values not supported.
脚本
import pandas as pd
def load_dataset():
ds = pd.read_csv('hour.csv')
ds['dteday'] = pd.to_datetime(ds['dteday'])
return ds
def one_hot_encoding(df, field):
one_hot_encoded = pd.get_dummies(df[field])
return pd.concat([df.drop(field, axis=1), one_hot_encoded], axis=1)
def preprocess_dataset(df):
df_reduced = df[['dteday', 'cnt', 'season','yr', 'mnth','hr', 'holiday', 'weekday', 'workingday', 'weathersit', 'temp', 'atemp', 'hum', 'windspeed']]
df_reduced = one_hot_encoding(df_reduced, 'season')
df_reduced = one_hot_encoding(df_reduced, 'mnth')
df_reduced = one_hot_encoding(df_reduced, 'hr')
df_reduced = one_hot_encoding(df_reduced, 'weekday')
df_reduced = one_hot_encoding(df_reduced, 'weathersit')
return df_reduced
dataset = load_dataset()
dataset = preprocess_dataset(dataset)
from datetime import datetime
def filter_by_date(ds, start_date, end_date):
start_date_parsed = datetime.strptime(start_date, "%Y-%m-%d")
start_end_parsed = datetime.strptime(end_date, "%Y-%m-%d")
return ds[(ds['dteday'] >= start_date_parsed) & (ds['dteday'] <= start_end_parsed)]
train = filter_by_date(dataset, '2011-01-01', '2012-10-31')
dev = filter_by_date(dataset, '2012-11-01', '2012-11-30')
val = filter_by_date(dataset, '2012-12-01', '2012-12-31')
import numpy as np
def reshape_dataset(ds):
Y = ds['cnt'].values
ds_values = ds.drop(['dteday', 'cnt'], axis=1).values
X = np.reshape(ds_values, (ds_values.shape[0], 1, ds_values.shape[1]))
return X, Y
X_train, Y_train = reshape_dataset(train)
X_dev, Y_dev = reshape_dataset(dev)
X_val, Y_val = reshape_dataset(val)
import keras
from matplotlib import pyplot as plt
from IPython.display import clear_output
class PlotLosses(keras.callbacks.Callback):
def on_train_begin(self, logs={}):
self.i = 0
self.x = []
self.losses = []
self.val_losses = []
self.fig = plt.figure()
self.logs = []
def on_epoch_end(self, epoch, logs={}):
self.logs.append(logs)
self.x.append(self.i)
self.losses.append(logs.get('loss'))
self.val_losses.append(logs.get('val_loss'))
self.i += 1
clear_output(wait=True)
plt.plot(self.x, self.losses, label="loss")
plt.plot(self.x, self.val_losses, label="val_loss")
plt.legend()
plt.show()
plot_losses = PlotLosses()
from keras.models import Model
from keras.layers import Input, Dense, LSTM, Dropout
def get_model():
input = Input(shape=(1, 58))
x = LSTM(200)(input)
x = Dropout(.5)(x)
activation = Dense(1, activation='linear')(x)
model = Model(inputs=input, outputs=activation)
optimizer = keras.optimizers.Adam(lr=0.01,
beta_1=0.9,
beta_2=0.999,
epsilon=None,
decay=0.001,
amsgrad=False)
model.compile(loss='mean_absolute_error', optimizer=optimizer)
model.summary()
return model
get_model()
def train_model(model, X_train, Y_train, validation, callbacks):
model.fit(X_train, Y_train, epochs=200, batch_size=1024, validation_data=validation, callbacks=callbacks, shuffle=False)
return model
model = train_model(get_model(), X_train, Y_train, (X_dev, Y_dev), [plot_losses])
数据集: Bike sharing dataset
期望退出
最佳答案
我在 Google Colab 中对您的脚本做了轻微的修改，直接从网络加载 zip 并对其进行处理(代码包含在下面)，并且我没有收到任何错误。不完全确定有什么不同，但这个版本可能有用 - 也许没有从本地 csv 正确读取拟合过程的输入数据 - 我希望这有帮助:
# Source for download_extract_zip:
# https://techoverflow.net/2018/01/16/downloading-reading-a-zip-file-in-memory-using-python/
from zipfile import ZipFile
import requests
import io
import zipfile
def download_extract_zip(url):
"""
Download a ZIP file and extract its contents in memory
yields (filename, file-like object) pairs
"""
response = requests.get(url)
with zipfile.ZipFile(io.BytesIO(response.content)) as thezip:
for zipinfo in thezip.infolist():
with thezip.open(zipinfo) as thefile:
yield zipinfo.filename, thefile
import pandas as pd
def load_dataset():
ds=''
raw_dataset = 'https://archive.ics.uci.edu/ml/machine-learning-databases/00275/Bike-Sharing-Dataset.zip'
for (iFilename, iFile) in download_extract_zip(raw_dataset):
if iFilename == 'hour.csv':
ds = pd.read_csv(iFile)
ds['dteday'] = pd.to_datetime(ds['dteday'])
return ds
def one_hot_encoding(df, field):
one_hot_encoded = pd.get_dummies(df[field])
return pd.concat([df.drop(field, axis=1), one_hot_encoded], axis=1)
def preprocess_dataset(df):
df_reduced = df[['dteday', 'cnt', 'season','yr', 'mnth','hr', 'holiday', 'weekday', 'workingday', 'weathersit', 'temp', 'atemp', 'hum', 'windspeed']]
df_reduced = one_hot_encoding(df_reduced, 'season')
df_reduced = one_hot_encoding(df_reduced, 'mnth')
df_reduced = one_hot_encoding(df_reduced, 'hr')
df_reduced = one_hot_encoding(df_reduced, 'weekday')
df_reduced = one_hot_encoding(df_reduced, 'weathersit')
return df_reduced
dataset = load_dataset()
dataset = preprocess_dataset(dataset)
from datetime import datetime
def filter_by_date(ds, start_date, end_date):
start_date_parsed = datetime.strptime(start_date, "%Y-%m-%d")
start_end_parsed = datetime.strptime(end_date, "%Y-%m-%d")
return ds[(ds['dteday'] >= start_date_parsed) & (ds['dteday'] <= start_end_parsed)]
train = filter_by_date(dataset, '2011-01-01', '2012-10-31')
dev = filter_by_date(dataset, '2012-11-01', '2012-11-30')
val = filter_by_date(dataset, '2012-12-01', '2012-12-31')
import numpy as np
def reshape_dataset(ds):
Y = ds['cnt'].values
ds_values = ds.drop(['dteday', 'cnt'], axis=1).values
X = np.reshape(ds_values, (ds_values.shape[0], 1, ds_values.shape[1]))
return X, Y
X_train, Y_train = reshape_dataset(train)
X_dev, Y_dev = reshape_dataset(dev)
X_val, Y_val = reshape_dataset(val)
import keras
from matplotlib import pyplot as plt
from IPython.display import clear_output
class PlotLosses(keras.callbacks.Callback):
def on_train_begin(self, logs={}):
self.i = 0
self.x = []
self.losses = []
self.val_losses = []
self.fig = plt.figure()
self.logs = []
def on_epoch_end(self, epoch, logs={}):
self.logs.append(logs)
self.x.append(self.i)
self.losses.append(logs.get('loss'))
self.val_losses.append(logs.get('val_loss'))
self.i += 1
clear_output(wait=True)
plt.plot(self.x, self.losses, label="loss")
plt.plot(self.x, self.val_losses, label="val_loss")
plt.legend()
plt.show()
plot_losses = PlotLosses()
from keras.models import Model
from keras.layers import Input, Dense, LSTM, Dropout
def get_model():
input = Input(shape=(1, 58))
x = LSTM(200)(input)
x = Dropout(.5)(x)
activation = Dense(1, activation='linear')(x)
model = Model(inputs=input, outputs=activation)
optimizer = keras.optimizers.Adam(lr=0.01,
beta_1=0.9,
beta_2=0.999,
epsilon=None,
decay=0.001,
amsgrad=False)
model.compile(loss='mean_absolute_error', optimizer=optimizer)
model.summary()
return model
get_model()
def train_model(model, X_train, Y_train, validation, callbacks):
model.fit(X_train, Y_train, epochs=200, batch_size=1024, validation_data=validation, callbacks=callbacks, shuffle=False)
return model
model = train_model(get_model(), X_train, Y_train, (X_dev, Y_dev), [plot_losses])
关于python - 使用 Keras 进行时间序列预测 - 模型值错误，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/58698942/