All Apps and Add-ons

Forecasting vs predicting (regression)



I am confused about Forecasting (ex: with Kalman filter) and Predicting (ex: with SGDRegressor or Logistic Regression).

I have a dataset with the number of logins by hour during 1 month (MAY). My variables are: _time and logins
I used SGDRegressor and saved the model (month of may). Then I applied the saved model with data of JUNE (this is unseen data).
Results are very good (I do not know if I am overfitting). But as I said. I am confused about Forecasting with Kalman and Predicting (univariate) with SGDRegressor.

Any help, please?

Here, I am using 36 hours as test and I am forecasting 36 hours

| inputlookup fortigate_QC_May2019_logins.csv
| fit StandardScaler "logins" with_mean=false with_std=true
| fit DBSCAN "SS_logins"
| where NOT cluster==-1
| predict "SS_logins" as prediction algorithm=LLP5 holdback=36 future_timespan=72 upper95=upper95 lower95=lower95 
| `forecastviz(72, 36, "SS_logins", 95)`

Here, I am calling a saved model (using MAY) and applying it with JUNE.
| inputlookup fortigate_QC_June2019_logins.csv
| fit StandardScaler "logins" with_mean=false with_std=true
| apply "authentication_profiling_SGDRegressor"
| table _time, "SS_logins", "predicted(SS_logins)"
0 Karma


Hi @rosho,

Maybe this explanation here might help:

Forecasting would be a subset of prediction. Any time you predict into the future it is a forecast. All forecasts are predictions, but not all predictions are forecasts, as when you would use regression to explain the relationship between two variables.

Hope this clears your confusion,

0 Karma
Did you miss .conf21 Virtual?

Good news! The event's keynotes and many of its breakout sessions are now available online, and still totally FREE!