使用 LTSM 进行时间序列预测，无需交叉验证

Question

运行超参数调整后，我有以下代码来绘制预测和实际值。

def cv(n_splits, LSTM_unit, lr, epoch, batch_size, saving_dir,
       X, Y, seed):
  """
  This is the function to do cross validation from the hyperparameter tuning 
  that we already have
  Args:
    :n_split:     Number of cv split(int)
    :LSTM_unit:   Number of unit in LSTM model(int)
    :lr:          Learning Rate for the model
    :epoch:       Epoch for the model(int)
    :batch_size:  Batch size for the model
    :saving_dir:  Directory in which the model is saved(str)
    :X:           Feature set (np.ndarray)
    :Y:           Target label (np.ndarray)
    :seed:        seed for CV random state(int)
  """
  model_name = " ".join([str(LSTM_unit), str(lr), str(epoch), str(batch_size)])
  kfold = KFold(n_splits=n_splits, shuffle=True, random_state=seed)
  cvscores = []
  hist_pred = []
  hist_actual = []
  # Load the model model
  model = load_model(saving_dir+"{}.h5".format(model_name))
  for train, test in kfold.split(X, Y):
  # evaluate the model
    X[test] = np.array(X[test])
    X[test] = X[test].reshape((X[test].shape[0], 1, X[test].shape[-1]))
    predictions = model.predict(X[test])
    # print(predictions) 
    hist_pred.extend(predictions)
    hist_actual.extend(Y[test])
    scores = np.sqrt(np.mean((predictions - Y[test]**2)))
    cvscores.append(scores)
  print('CV-RMSE = {} with {} standard deviation'.format(np.mean(cvscores), 
                                                         np.std(cvscores)))


  hist_pred = label_scaler.inverse_transform(hist_pred)
  hist_actual = label_scaler.inverse_transform(hist_actual)

  #Plot/Create the data for the graph
  plt.figure(figsize=(10,5))
  plt.plot(hist_pred, color = 'green', label = 'Predictions')
  plt.plot(hist_actual, color = 'red', label = 'Real Value')
  plt.title('')

  plt.ylabel('')
  plt.legend()
  plt.show()

代码运行良好。我想要做的是，使用上面的代码（没有交叉验证）进行样本预测。我的意思是，想象一个时间序列数据集，它有特征变量但没有标签。我想在没有交叉验证的情况下使用已经建立的模型来预测这些标签。既然是时间序列未来预测，用CV没有意义。

“举个例子，假设我想预测本月底将发布的通货膨胀。我现在拥有所有特征变量来预测它。”

如何修改上面的代码？任何帮助将不胜感激。

谢谢。