Predict time series data by group

Asked 2 weeks ago, Updated 2 weeks ago, 1 views

Hello, I am studying machine learning. There are time series data as below.

time                   id         value
2010-04-11 23:00:00 2156213     10.0    
2010-04-12 02:00:00 2156213     3.0
2010-04-12 08:00:00 2156213     9.0
2010-02-01 04:00:00 2231441     0.0
2010-02-01 11:00:00 2431441     7.0
... ... ... ... ... ... ...
2016-10-13 08:00:00 2441232     134.0
2016-10-13 10:00:00 2441232     136.0
2016-10-13 14:00:00 2441232     140.0
2016-10-13 16:00:00 2441232     100.0
2016-10-14 04:00:00 2441232     120.0

After learning the time series data above, I want to predict the next value given the new id, time, and value. (The measured time, interval, and value change for each id are different.)

I thought of a way to combine the learned models after conducting regression learning separately for each ID, but there seems to be no way to combine them. I don't think there's an algorithm like that, but I don't know if I'm thinking too hard. Does anyone know any other way?

machine-learning

2022-09-20 12:33

1 Answers

Usually, the time series is based on rnn, lstm, etc. I wonder if the data you asked can be applied because the time interval is not constant.

Regression is... It's a prediction of the relationship between an independent variable and a dependent variable, is there anything that can be taken as an independent variable? Date time is Monday, Sunday, day, hour, minute, second... Should I do something with my back? I don't think I can guarantee the result even if I do this.


2022-09-20 12:33

If you have any answers or tips


© 2022 pinfo. All rights reserved.