我正在研究一个序列预测问题，我在这方面没有太多经验，所以下面的一些问题可能很幼稚。
仅供参考：我在这里创建了一个关注 CRF 的后续问题
我有以下问题：
我想预测多个非独立变量的二进制序列。
输入：
我有一个包含以下变量的数据集：
- 时间戳
- A组和B组
- 在特定时间戳对应每个组的二进制信号
此外，假设以下情况：
- 我们可以从时间戳（例如一天中的小时）中提取额外的属性，这些属性可以用作外部预测器
- 我们认为 A 组和 B 组不是独立的，因此联合建模他们的行为可能是最佳的
binary_signal_group_A并且binary_signal_group_B是我想使用（1）它们过去的行为和（2）从每个时间戳中提取的附加信息来预测的 2 个非独立变量。
到目前为止我所做的：
# required libraries
import re
import numpy as np
import pandas as pd
from keras import Sequential
from keras.layers import LSTM
data_length = 18 # how long our data series will be
shift_length = 3 # how long of a sequence do we want
df = (pd.DataFrame # create a sample dataframe
.from_records(np.random.randint(2, size=[data_length, 3]))
.rename(columns={0:'a', 1:'b', 2:'extra'}))
# NOTE: the 'extra' variable refers to a generic predictor such as for example 'is_weekend' indicator, it doesn't really matter what it is
# shift so that our sequences are in rows (assuming data is sorted already)
colrange = df.columns
shift_range = [_ for _ in range(-shift_length, shift_length+1) if _ != 0]
for c in colrange:
for s in shift_range:
if not (c == 'extra' and s > 0):
charge = 'next' if s > 0 else 'last' # 'next' variables is what we want to predict
formatted_s = '{0:02d}'.format(abs(s))
new_var = '{var}_{charge}_{n}'.format(var=c, charge=charge, n=formatted_s)
df[new_var] = df[c].shift(s)
# drop unnecessary variables and trim missings generated by the shift operation
df.dropna(axis=0, inplace=True)
df.drop(colrange, axis=1, inplace=True)
df = df.astype(int)
df.head() # check it out
# a_last_03 a_last_02 ... extra_last_02 extra_last_01
# 3 0 1 ... 0 1
# 4 1 0 ... 0 0
# 5 0 1 ... 1 0
# 6 0 0 ... 0 1
# 7 0 0 ... 1 0
# [5 rows x 15 columns]
# separate predictors and response
response_df_dict = {}
for g in ['a','b']:
response_df_dict[g] = df[[c for c in df.columns if 'next' in c and g in c]]
# reformat for LSTM
# the response for every row is a matrix with depth of 2 (the number of groups) and width = shift_length
# the predictors are of the same dimensions except the depth is not 2 but the number of predictors that we have
response_array_list = []
col_prefix = set([re.sub('_\d+$','',c) for c in df.columns if 'next' not in c])
for c in col_prefix:
current_array = df[[z for z in df.columns if z.startswith(c)]].values
response_array_list.append(current_array)
# reshape into samples (1), time stamps (2) and channels/variables (0)
response_array = np.array([response_df_dict['a'].values,response_df_dict['b'].values])
response_array = np.reshape(response_array, (response_array.shape[1], response_array.shape[2], response_array.shape[0]))
predictor_array = np.array(response_array_list)
predictor_array = np.reshape(predictor_array, (predictor_array.shape[1], predictor_array.shape[2], predictor_array.shape[0]))
# feed into the model
model = Sequential()
model.add(LSTM(8, input_shape=(predictor_array.shape[1],predictor_array.shape[2]), return_sequences=True)) # the number of neurons here can be anything
model.add(LSTM(2, return_sequences=True)) # should I use an activation function here? the number of neurons here must be equal to the # of groups we are predicting
model.summary()
# _________________________________________________________________
# Layer (type) Output Shape Param #
# =================================================================
# lstm_62 (LSTM) (None, 3, 8) 384
# _________________________________________________________________
# lstm_63 (LSTM) (None, 3, 2) 88
# =================================================================
# Total params: 472
# Trainable params: 472
# Non-trainable params: 0
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy']) # is it valid to use crossentropy and accuracy as metric?
model.fit(predictor_array, response_array, epochs=10, batch_size=1)
model_preds = model.predict_classes(predictor_array) # not gonna worry about train/test split here
model_preds.shape # should return (12, 3, 2) or (# of records, # of timestamps, # of groups which are a and b)
# (12, 3)
model_preds
# array([[1, 0, 0],
# [0, 0, 0],
# [1, 0, 0],
# [0, 0, 0],
# [1, 0, 0],
# [0, 0, 0],
# [0, 0, 0],
# [0, 0, 0],
# [0, 0, 0],
# [0, 0, 0],
# [1, 0, 0],
# [0, 0, 0]])
问题：
这里的主要问题是：我如何让它工作，以便模型预测两组的下一个 N 序列？
此外，我想问以下问题：
- 预计 A 组和 B 组是互相关的，但是，尝试通过单个模型同时输出 A 和 B 序列是否有效，或者我应该拟合 2 个单独的模型，一个预测 A，另一个预测 B，但都使用历史 A 和 B 数据作为输入？
- 虽然我在模型中的最后一层是形状为 (None, 3, 2) 的 LSTM，但预测输出的形状为 (12, 3)，而我预计它是 (12, 2) - 我在做什么这里错了，如果是这样，我将如何解决这个问题？
- 就输出 LSTM 层而言，在这里使用激活函数会不会是个好主意，比如 sigmoid？为什么/为什么不？
- 使用分类类型损失（二进制交叉熵）和度量（准确度）来优化序列是否有效？
- LSTM 模型在这里是最佳选择吗？有人认为 CRF 或某些 HMM 类型的模型在这里会更好吗？
非常感谢！