Demantra Analytical Engine Tuning
Hi all,
We are on a Demantra 7.2.0.2 version.
I loaded 1 year of historical data (weekly time bucket) and I made a forecast on the next year using:
-all available Models except Holt,
-need_spread - produce forecast with spikes,
-test_samp_len - 52
and besides that all system parameters are default.
The problem is that when I checked the results I had a very good historical fit like Abs % Error - between 1% - 7%
and the forecast was almost a straight line which is not the case because the sales data is very "jumpy".
So if anyone has any ideea about what parameters I should check I will be very glad.
We are on a Demantra 7.2.0.2 version.
I loaded 1 year of historical data (weekly time bucket) and I made a forecast on the next year using:
-all available Models except Holt,
-need_spread - produce forecast with spikes,
-test_samp_len - 52
and besides that all system parameters are default.
The problem is that when I checked the results I had a very good historical fit like Abs % Error - between 1% - 7%
and the forecast was almost a straight line which is not the case because the sales data is very "jumpy".
So if anyone has any ideea about what parameters I should check I will be very glad.
0