为什么将原代码中--features改为MS后，预测的结果全是负的零点几呢？

[sfk123]

下面是我预测出来的结果文本：

date,data 2018-06-26 20:00:00,-0.573083 2018-06-26 21:00:00,-0.4734673 2018-06-26 22:00:00,-0.50346094 2018-06-26 23:00:00,-0.5455254 2018-06-27 00:00:00,-0.5396262 2018-06-27 01:00:00,-0.54037297 2018-06-27 02:00:00,-0.57873106 2018-06-27 03:00:00,-0.41970533 2018-06-27 04:00:00,-0.36646178 2018-06-27 05:00:00,-0.48071027 2018-06-27 06:00:00,-0.39691973 2018-06-27 07:00:00,-0.35621166 2018-06-27 08:00:00,-0.37568408 2018-06-27 09:00:00,-0.38776475 2018-06-27 10:00:00,-0.39422265 2018-06-27 11:00:00,-0.3724526 2018-06-27 12:00:00,-0.25835124 2018-06-27 13:00:00,-0.30427188 2018-06-27 14:00:00,-0.32799682 2018-06-27 15:00:00,-0.378375 2018-06-27 16:00:00,-0.26541272 2018-06-27 17:00:00,-0.30582866 2018-06-27 18:00:00,-0.28742903 2018-06-27 19:00:00,-0.27987918

[faisal3325]

The below line uses the last column of the data for prediction target. So make sure to have your target column placed last in the dataframe.

batch_y = batch_y[:,-self.args.pred_len:,f_dim:].to(self.device)

exp/exp_informer -> Line 287