Time series prediction

[anebz]

Until now, the model I've used for prediction has been regression. But our data resembles a time series approach more, where each interval lasts one day. The prediction for the occupancy today at 17:00 in the Frankfurt gym depends more on the occupancy today at 16:45 in the Frankfurt gym, than in the average occupancy at 17:00 in the Frankfurt gym over previous days and weeks.

I have no idea about time series, so this is a freestyle issue. Ideally I'd like to have a time series model trained with our data, with all the preprocessing it needs, and a way that we can call the trained model and get a prediction for a new entry.

[anebz]

Theory: https://otexts.com/fpp2/judgmental-principles.html
streamlit prophet: https://share.streamlit.io/maximelutel/streamlit_prophet/main/streamlit_prophet/app/dashboard.py
prophet: https://facebook.github.io/prophet/