1
$\begingroup$

Is time series model suitable for network syslogs considering the fact the messages are sequential and the messages are outputted as a result of dependency between themselves which can range from normal notifications to alarms if there are.

Can the state of the art models like ARIMA be applied ?

What kind of questions can be answered with applying time series models on syslogs ?

$\endgroup$

1 Answer 1

0
$\begingroup$

No, according to your description the syslog is not a time series data.

Time series data means to have a numerical value at each time stamps. For instance, if your log records the CPU usage (a number between 0 to 100%) then it is a time series data.

If you log is a sequence of text message, like the /var/log/syslog in Linux, then it is not a time series data.

$\endgroup$
1
  • $\begingroup$ Thank you for this! But what if the messages are transformed into their respective frequency counts, say hourly; would it be appropriate to use time series model? If not, so what kind of models can be used in forecasting with syslogs ? $\endgroup$
    – Bode
    Sep 22, 2018 at 6:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.