The challenge is to predict time series (y) from a very similar univariate time series (x) without using the identify function (y=x). There are ~73,000 timestamps of training with one measurement every 5 minutes.
The following image shows a subsample of the training time series:

![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1150148%2Ffa517edca69b408f4fff1f703aa16b90%2Fnewplot.png?generation=1683915723349979&alt=media)