Data Preprocessing Issue

1

sample_data
Dear all,

Good day. I am new to machine learning.
May I get some advice on data preprocessing?
The attached image is sample data, which the first column is time step, second column is item_ID, and the last column is speed.

My question is, how can I create a sequences of time series based on each item_ID?
I don’t want the the sequences is mix of both item.
I am aware of creating cyclic sequences. But the length of time step for each item is different.

May I get some advice from you all?

Thank you very much.

Regards,
GMC

2

Hmmm it’s a long story.
When data is different length you can pad the data.
Nowadays, there are architectures which deals with different-length samples (transformers).

The way to go is measuring or estimating the maximum length possible and then padding your samples according to that.

Another way is doing chunks of a representative length and working with a reduced duration.