Building Timeseries models for stock trading having multiple stocks

External question — Data Science Stack Exchange Author: Ravi Original post: https://datascience.stackexchange.com/questions/76136 License: CC BY-SA 4.0 — https://creativecommons.org/licenses/by-sa/4.0/ Adaptation: HTML converted to plain text; contact email addresses removed. I have gone through some of the tutorials on the timeseries and all of them have taken one stock for the timeseries and tried to forecast it. My dataset contains many stocks for the time period(each day). Do i need to choose one stock for or some of them or all of them for training the model?

External answer — Data Science Stack Exchange Author: babelproofreader Original post: https://datascience.stackexchange.com/a/76141 License: CC BY-SA 4.0 — https://creativecommons.org/licenses/by-sa/4.0/ Adaptation: HTML converted to plain text; contact email addresses removed. This is a question about Univariate versus Multivariate Time Series analysis. Both can be valid approaches. However, for stock price forecasting I would suggest that Multivariate will probably be better, with the important caveat that your choice of time series should have some fundamental justification. Randomly choosing to analyse stock A and B together from a pool of 'N' thousands of potential stock pairs/triplets etc. will not be as effective as choosing stock X and Y if you believe X and Y are related in some way, i.e. they are in the same industry sector.