Level Id in the mdp_matrix ( Oracle Demantra)

Question

The level id in the mdp_matrix signifies the level at which the forecast gets generated for a given item-location combination.
I have six levels defined in the forecast tree.
When I see the results for level id in mdp_matrix after having run the analytical engine I am seeing values of -1, 0 and 6 - what do they signify.
Which table stores these level id and equivalent forecast tree level.
I am bringing sales historical data at the lowest level - for both item and ship to code.
I am atleast expecting some/few/most of the forecast should be generated at the lowest level - level 1 (level_id) - but hardly do I find records in the mdp_matrix as level_id =1 (the min level in the forecast model library is set to 2 / 3 and max level to 999999.
Any pointers appreciated.
thx and rgds,
Pankaj

Answer 1

Hi Pankaj,
The level_id field is updated dynamically when you run the analytical engine. The range of values in level_id field is determined by number of forecast levels you choose in forecast tree definition. But while generating forecast, which values to use among is determined by max_fore_level and min_fore_level parameters as mentioned by Rubellah.
Take the following example: (This is the log generated by Analytical Engine, stored in "engine2k" folder)
In this log statement [x,y,z] values signify, x - level_id, y - item_node and z - loc_node respectively.
The Analytical engine analyses every set of data available for forecast for each item-location combination. But always generate forecast at lowest possible level.
1) 10:36:44:967 [3,22,1] has a usable number of observations. (52)
2) 10:36:44:998 [2,22,1] has a usable number of observations. (52)
3) 10:36:45:280 [2,22,1] forecast succeeded using models B,C,E,X
4) 10:37:24:219 [3,53,8] has a usable number of observations. (52)
5) 10:37:24:265 [2,53,8] has a usable number of observations. (52)
6) 10:37:24:922 [2,53,8] forecast failed.
7) 10:37:25:578 [3,53,8] forecast failed.
8) 10:37:25:797 [3,55,8] has a usable number of observations. (4)
9) 10:37:25:828 [2,55,8] has a usable number of observations. (4)
10) 10:37:25:844 [2,55,8] forecast failed.
11) 10:37:25:859 [3,55,8] forecast failed.
12) 10:37:26:047 [3,57,8] has a usable number of observations. (51)
13) 10:37:26:078 [2,57,8] has a usable number of observations. (51)
14) 10:37:26:078 Node has Intermittent data.
15) 10:37:26:125 [2,57,8] forecast succeeded using models F
16) 10:37:26:906 [4,3,8] forecast succeeded using models B,C,E
17) 10:37:27:250 [4,4,8] has a usable number of observations. (52)
As you see line 1, shows that there is usable no of observations it found item_node 22 & loc_node 1 combination at higher level_id 3. But according to line 2 it also found good enough observations at lower level_id 2. Hence it generated the forecast at level_id 2 as sighted in the 3rd line.
The exceptions might occur like in lines 4,5,6 & 7 where engine is not able to generate forecast at lower as well as upper level too. The first try is made at lower level [2,53,8](It failed) the 2nd try made at upper level [3,53,8](It failed here too).
From the third set of lines 8,9,10 & 11 we can guess the reason of forecast being failed as very less number of observations (only 4).
The forth set of lines 12,13,14 & 15 shows that there is intermittent data available so it generated forecast at lower level using model "F".
Line 16 states that forecast is generated at upper level_id 4.
Line 17 states that there were usable no of observations but didn't generate any forecast for that combination.
Hope this might help you.
Milind...