Ordering results of Association mode prediction by Suppport
- Hello,
Below is the query I use to predict associations and I want to order by Support. However the query seems to order by Adjusted Probability.
It seems that the default behaviour of Predict is to order by $probabilty or Adjusted Probabilty.
Can someone suggest how to overcome this(while retaining the node_id<>'' condition)
SELECT
FLATTENED
(
SELECT * FROM Predict([Recommendation],INCLUDE_STATISTICS,$Suppport,5,INCLUDE_NODE_ID)
WHERE $NODEID<>'' )
From [Recommendation_Model]
NATURAL
PREDICTION JOIN (SELECT ( Select 'xxxxxx' AS [Table Nm])
AS [Recommendation]) AS t
Please treat this urgent!
Thanks
Kirthika
All Replies
- Would greatly appreciate if someone can help answer this question.-Kirthika
- You can try using TopCount() function and selecting the column to order by. For more details, see the following:
http://msdn.microsoft.com/en-us/library/ms132149.aspx
Hope this helps
Shuvro
MSFT, SQL Server Data Mining- Proposed As Answer byJamie MacLennanMSFT, ModeratorMonday, October 19, 2009 6:49 PM
- Hello Shuvro,Thank you for your response. But I am not using TopCount as I do not want a fixed number of results. I need results ordered by $Support if they exists (if rules exists)This is what Jamie explains in http://blogs.msdn.com/jamiemac/archive/2006/04/19/predict-based-on-rules-alone.aspxfurther, what I do not understand is as per msdn Predict () is supposed to predict based on $Support or $Probability based on what is specified in the query. But that is not happeninig in the above query. Can you help explain the reason behind this ?If this is a known issue, can you please suggest an alternative to select based on rules and still order by $Support.(I am not sure how to order by $support)ThanksKirthika
- Hello,
It will be really helpful if someone on the forum can help reply to the above .
Thanks
Kirthika - Are you using SQL Server 2005 or 2008?
I tried on a couple of models and got the expected results, sorted correctly, by $SUPPORT, in descending order.
Could you please post the results returned by the following query?
SELECT FLATTENED Predict([Recommendation], INCLUDE_STATISTICS, $Support, 5, INCLUDE_NODE_ID)
From [Recommendation_Model]
NATURAL PREDICTION JOIN (SELECT ( Select 'xxxxxx' AS [Table Nm]) AS Recommendation AS T
bogdan crivat [sql server data mining] / http://www.bogdancrivat.net/dm Hello Bogdan,
Thank you for your response. I am using SQL server AS 08.
The query above gives me the following result. Note that ordering by support here says that 3106 is the support of the top item.
XXXXX 3106 0.399557637262555 0.815711212151828 801
YYYYY 15 0.00182149362477231 0.811337767743637 801
ZZZZZ 7 0.000910746812386157 0.8165614747249 801
WWW 2 0.000260213374967473 0.8165614747249 801
UUUU 1 0.000130106687483737 0.8165614747249 801
Now if I use the same query with an ‘n’ value of 30 I am getting different support level infact higher ones which did not appear in the result above. This makes me think that the query is NOT ordering by descending order of support. Am I missing something here ?
AAAA 7454 0.702706219099662 0.791194273501276 801
BBBB 5352 0.500780640124902 0.79057488043516 801
CCCC 4464 0.49687743950039 0.804583927418361 801
DDDD 4434 0.4463960447567 0.796570710182286 801
EEEEE 4385 0.481134530314858 0.803454066489495 801
FFFFF 3448 0.418293000260213 0.811263202473559 801
XXXXX 3106 0.399557637262555 0.815711212151828 801
HHHH 1387 0.103434816549571 0.771172642705494 801
IIIIIIII 1217 0.102654176424668 0.781859160343093 801
JJJJJ 433 0.0416341399947957 0.792820986466476 801
Thanks
KirthikaI am not able to repro the behavior you are describing It is also strange that all results have the same NODE_ID value.
Could you share the model? EXPORT the model without data and send it by email to bogdanc AT microsoft DOT com?
bogdan crivat [sql server data mining] / http://www.bogdancrivat.net/dm- Hello Bogdan,
Thanks for looking into this. I will export and send the model to you.
Thanks
Kirthika


