CODEWITHSUNDEEP
algorithmrtime-series
Avinash
Avinash

Reputation: 13267

forecasting resource utilization

I want to create a system to forecast certain resource utilization; for example, CPU utilization. I have data of CPU utilization for each day. How can I predict its usage for next future time, say 2 days? I know that time series analysis can help but I fail to understand how to accommodate other factors associated with the CPU utilization as time series analysis is only time on x-axis and utilization on y-axis.

Upvotes: 0

Views: 1280

Answers (2)

ECII
ECII

Reputation: 10629

Check this out, i think it can help you a lot or at least help you start with something. He deals with a similar problem (forecasting of hard disk space requirements)

http://lpenz.github.com/articles/df0pred-1/index.html

http://lpenz.github.com/articles/df0pred-2/index.html

http://lpenz.github.com/articles/df0pred-3/index.html

Upvotes: 1

pbarill
pbarill

Reputation: 680

I deduce that you have multiple time series, and that you want to put this extra information at work (as opposed to a univariate model solely with cpu utilization).

For a univariate model, you can check with arima(), and find a suitable order for this model using auto.arima() in package forecast. Predictions can be made using predict(), on the arima object.

For a multivariate model, you can consider a vector auto-regressive model. Check for function VAR() in package vars.

Upvotes: 0

Related Questions