Hi,

I am finding difficulty understand how to learn from the provided data. I will first present my understanding of the dataset then give my question.

In similar datasets, such as the ETTh dataset, you have the rows referring to the consecutive timesteps. Then the columns referring to different features, then, and importantly, the last columns tagged “OT” representing the label or the target value.

In the provided dataset, things are a bit different. There are 4 files, one of them, named “traffic_DLThpVol.csv” represent the target values and the other 3 files represent 3 features. Every column in these csv files (2,880 columns in total) represent a different scenario (specifically, we have 30 base stations, each one has 3 cells and each cell transmit 32 beams, so in total, 30x3x32=2,880).

The objective is to forecast the values of the throughput value as provided in the “traffic_DLThpVol.csv” for some period of time in the future (168 rows x 2,880 columns).

Could anyone help me understand how I should pass these data to the learning algorithm? I am trying to use the iTransformer algorithm first from their github repo.

Thanks,