Time Series features Encoding for Sequence Data Problem

I have time data (2011-06-09T23:29:30Z) in the dataset. I am giving these data as input to the LSTM in the Deep Learning model.

I did some research on how to best encode this time data and convert it to an embedding vector.

I saw an example in the Python original document and applied it

I guess they call this embedding method the Epoch time format

When I apply it to all time data, I get a result like the one below.

I have nearly 1 million data, this is just a sample piece.

but after applying this, I wondered if there is a more efficient embedding method for LSTM entry and continued to search.

It is said that the time representation method should be chosen according to the problem.

My problem is that there are users in the dataset and they visit some locations, their visiting time is my time data column.

I would like to ask people who used time data in Deep Learning or Machine Learning models before. What is the best time representation solution for me.

from Time Series features Encoding for Sequence Data Problem