Hey all,
I am currently trying to achieve the following:
train a Kalman filter with a periodicity i found via Autocorrelation on the last 3 weeks data and make prediction for one week of future data. I do this as follows:
index = cisco_prod
| timechart span=1h count as logins_hour
| fit ACF logins_hour k=200 fft=true conf_interval=95 as corr
| top limit=2 acf(corr),Lag
| stats max(Lag) as corr_lag
| map search="search index = cisco_prod | timechart span=1h count as logins_hour | predict \"logins_hour\" as prediction algorithm=LLP holdback=200 future_timespan=368 period=$corr_lag$ upper95=upper95 lower95=lower95"
| `forecastviz(368, 200, "logins_hour", 95)`
But how do I now use this predictions for the coming week, to actually compare them to the incoming data? The thing is, I don't want to always train the Kalman filter with new data because if I feed it with anomalies it will not make correct predictions for the future.
Has anyone an idea?