Python Keras LSTM“检查输入时出错”

Question

我正在尝试训练 LSTM RNN。这是我第一次尝试这个并使用 Keras。我使用指南来编写此代码。

以下是数据示例：

    lat         long        datetime                id  trip_id    mode_cat
0   39.979973   116.305745  2011-08-27 06:13:01     20  1          1
1   39.979957   116.305688  2011-08-27 06:13:02     20  1          1
2   39.979960   116.305693  2011-08-27 06:13:03     20  1          1
3   39.979970   116.305717  2011-08-27 06:13:04     20  1          1
4   39.979985   116.305732  2011-08-27 06:13:05     20  1          1

其中 lat、long、datetime、id、trip_id 是输入，mode_cat 是输出（输出为二进制）。

这是我创建和训练 RNN 的部分代码：

def evaluate_model(trainX, trainy, testX, testy):
    verbose, epochs, batch_size = 0, 15, 64
    n_timesteps, n_features, n_outputs = trainX.shape[0], trainX.shape[1], 1
    model = Sequential()
    print("adding LSTM")
    model.add(LSTM(1, input_shape=(n_timesteps,n_features), return_sequences=True))
    print("adding dropout")
    model.add(Dropout(0.5))
    print("adding dense")
    model.add(Dense(100, activation='relu'))
    print("adding dense")
    model.add(Dense(n_outputs, activation='softmax'))
    print("adding compile")
    model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
    print(model.summary())
    # fit network
    model.fit(trainX, trainy, epochs=epochs, batch_size=batch_size, verbose=verbose)
    # evaluate model
    _, accuracy = model.evaluate(testX, testy, batch_size=batch_size, verbose=0)
    return accuracy

def run_experiment(repeats=10):
    # repeat experiment
    scores = list()
    for r in range(repeats):
        score = evaluate_model(train_x, train_y, test_x, test_y)
        score = score * 100.0
        print('>#%d: %.3f' % (r+1, score))
        scores.append(score)
    # summarize results
    summarize_results(scores)

run_experiment()

我得到以下结果：

adding LSTM
adding dropout
adding dense
adding dense
adding compile
Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
lstm_1 (LSTM)                (None, 455414, 1)         28        
_________________________________________________________________
dropout_1 (Dropout)          (None, 455414, 1)         0         
_________________________________________________________________
dense_1 (Dense)              (None, 455414, 100)       200       
_________________________________________________________________
dense_2 (Dense)              (None, 455414, 1)         101       
=================================================================
Total params: 329
Trainable params: 329
Non-trainable params: 0
_________________________________________________________________
None
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-8-5d4acc2b23d8> in <module>()
     31     summarize_results(scores)
     32 
---> 33 run_experiment()

<ipython-input-8-5d4acc2b23d8> in run_experiment(repeats)
     24     scores = list()
     25     for r in range(repeats):
---> 26         score = evaluate_model(train_x, train_y, test_x, test_y)
     27         score = score * 100.0
     28         print('>#%d: %.3f' % (r+1, score))

<ipython-input-8-5d4acc2b23d8> in evaluate_model(trainX, trainy, testX, testy)
     15     print(model.summary())
     16     # fit network
---> 17     model.fit(trainX, trainy, epochs=epochs, batch_size=batch_size, verbose=verbose)
     18     # evaluate model
     19     _, accuracy = model.evaluate(testX, testy, batch_size=batch_size, verbose=0)

~\Anaconda3\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, validation_freq, max_queue_size, workers, use_multiprocessing, **kwargs)
   1152             sample_weight=sample_weight,
   1153             class_weight=class_weight,
-> 1154             batch_size=batch_size)
   1155 
   1156         # Prepare validation data.

~\Anaconda3\lib\site-packages\keras\engine\training.py in _standardize_user_data(self, x, y, sample_weight, class_weight, check_array_lengths, batch_size)
    577             feed_input_shapes,
    578             check_batch_axis=False,  # Don't enforce the batch size.
--> 579             exception_prefix='input')
    580 
    581         if y is not None:

~\Anaconda3\lib\site-packages\keras\engine\training_utils.py in standardize_input_data(data, names, shapes, check_batch_axis, exception_prefix)
    133                         ': expected ' + names[i] + ' to have ' +
    134                         str(len(shape)) + ' dimensions, but got array '
--> 135                         'with shape ' + str(data_shape))
    136                 if not check_batch_axis:
    137                     data_shape = data_shape[1:]

ValueError: Error when checking input: expected lstm_1_input to have 3 dimensions, but got array with shape (455414, 5)

请指教！

Answer 1

馈入 RNN 与其他网络有点不同，因为它使用序列。

代码中的问题是数据集格式。 RNN 的输入应该是一个 3D 张量，形状为[batch, timesteps, feature]1

由于您有时间序列数据集，因此您应该使用移动窗口方案对输入数据进行预处理

您应该查看this tutorial 关于时间序列预测的信息，他们在其中实施了这样的窗口方案：

def multivariate_data(dataset, target, start_index, end_index, history_size,
                      target_size, step, single_step=False):
  data = []
  labels = []

  start_index = start_index + history_size
  if end_index is None:
    end_index = len(dataset) - target_size

  for i in range(start_index, end_index):
    indices = range(i-history_size, i, step)
    data.append(dataset[indices])

    if single_step:
      labels.append(target[i+target_size])
    else:
      labels.append(target[i:i+target_size])

  return np.array(data), np.array(labels)

past_history = 720
future_target = 72
STEP = 6

x_train_single, y_train_single = multivariate_data(dataset, dataset[:, 1], 0,
                                                   TRAIN_SPLIT, past_history,
                                                   future_target, STEP,
                                                   single_step=True)
x_val_single, y_val_single = multivariate_data(dataset, dataset[:, 1],
                                               TRAIN_SPLIT, None, past_history,
                                               future_target, STEP,
                                               single_step=True)

这里是移动窗口过程的示意图（取自paper）

您的问题不完全是时间序列预测，因此您应该根据您的问题调整该教程中的代码。 应该是这样的：

def moving_window(dataset, past_history):
    data, labels = [], []
    for i in range(past_history, len(dataset)):
        indices = range(i-past_history, i)
        data.append(dataset[['lat', 'long', 'id', 'trip_id']].values[indices])
        label.append(dataset['mode_cat'][i])
    return np.array(data), np.array(labels)

PAST_HISTORY = 60
X, y = moving_window(dataset, PAST_HISTORY)

PAST_HISTORY 是一个应该调整的超参数。

希望对您有所帮助！ :)