analysis service 2005: microsoft association rules problem

hi there,

i'm new forum .. maybe way of expression not good, hope understandable.

 

i've sql server 2005 database 90 columns , more or less 185 thousands records. i've run microsoft associations rules on laptop (sony vaio sz3, core 2 duo, 2gb ram).

the problems amount of ram seems not enough [it starts swap when it's reading 240th case)

 

because of this, decided sample data by extracting 10thousands records randomnly ... it lasts 25minutes (more or less) now, it's still much...

 

does better way exists? what's problem: column or row numbers?

 

hope guys can me.

the problem association rules parameters set on algorithm.

 

try - first create structure/model , deploy without processing.  process structure verify there isn't problem accessing data. 

 

if works fine, set model parameter minimum_support high - 0.80.  means items occur in 80% of cases counted.  since have dense data issue.   can gradually lower minimum support see how impacts performance , results.  additionally, since have dense data (e.g. attributes on single row) possible have data isn't interesting because true or everyone.  can set maximum_support number lower 1, example setting 0.9 eliminate items , itemsets occur in 90% or more of cases.

 

hth

-jamie

SQL Server  >  Data Mining