confident of data mining

in associate rule, there's confident and support or probabilty values. but i use sql 2000, no support associate rule so i create a model with Microsoft _decision _tree to predict other products when customer choose product in superMarket

basket_id

table...

{

productid

}

is it have confident or support values. what value of it? how can i change the minimum_support and minimum_probabilty ? it can use in Microsoft _ decision tree.

Hi

In the decision trees model, you could filter the prediction results by probability and obtain pretty much the same effect as in association rules.

To some extent, I think the minimum support from association rules can be modeled with the MINIMUM_LEAF_CASES parameter of the decision trees algorithm.

The difference comes from the "greedy" behavior of DT.

Assuming, in Association Rules, a minimum support of 10 and exactly 11 transaction including milk, beer and chips, a rule could be found like below:

"MILK + BEER => CHIPS"

Decision trees with a MINIMUM_LEAF_CASES of 10 will pick the most significant attribute ( in this case, either one of MILK and BEER, say MILK) and split on that attribute. After the split, there are not necessarily sufficient cases to support a new split on the second attribute (BEER), so the rule may be simpler:

IF (MILK) then CHIPS

Hope this helps

|||thanks, but i want ask default value of support value? what's it? and how i can modeled the MINIMUM_LEAF_CASES in decision trees algorithm ( i use interface not write code)|||I believe the default is 10. You can set parameters using the advanced properties of the mining model in the designer. Alternatively, I wrote (a long time ago) an addin for Analysis Manager that provides a UI for setting parameters in SQL 2000 data mining similar to that in SQL 2005. You can still get this addin here: http://www.sqlserverdatamining.com/DMCommunity/SQLServer2000/Links_LinkRedirector.aspx?id=97|||

thanks. i was try in advanced properties but fields is grey so i cant's modify.

10 is support. how about confident? has it?

i use link to get an addin you said, but it error.

so i haven't try it now.

|||

This is the direct link to the file mentioned by Jamie.http://www.sqlserverdatamining.com/DMCommunity/_Downloads/Downloads_GetFile.aspx?id=92

Note that it still requires creating an account on the sqlserverdatamining.com web site.

For confidence, you should filter the query results rather than looking for an algorithm property.

|||

i had download AddIn at http://www.sqlserverdatamining.com/DMCommunity/SQLServer2000/Links_LinkRedirector.aspx?id=97

and read document that requid you must regist file dll and registry. after, right click on model dataming you want to modify param and choose menu to modify it. But i can't do. how can i registry. i tryed double click on it ( file regist and dll, copy them to root folder) but nothing work, so i still can't modify param. how?

more: confider, you say me filter in query. how can i do that? can you give me an example query

|||

Open a command prompt, cd to the directory where the dll is and type

regsvr32 <dllname>

The .reg file should register itself when you double-click it.