The Data Mining Forum 
IMPORTANT: This is the old Data Mining forum.
I keep it online so that you can read the old messages.
Please post your new messages in the new forum: https://forum2.philippe-fournier-viger.com/index.php
Interestingness measure for closed itemsets
Posted by:
Vathsala
Date: April 22, 2014 10:02PM
Hi...
I have derived association rules from closed itemsets. The problem is while calculating confidence for rules derived from closed itemsets - if the anticident is not a closed itemset then anticident support cannot be fetched and hence the confidence is infinity... what can be done here or is there any other interestingness measure for rules derived from clodsed itemsets.
Regards
I have derived association rules from closed itemsets. The problem is while calculating confidence for rules derived from closed itemsets - if the anticident is not a closed itemset then anticident support cannot be fetched and hence the confidence is infinity... what can be done here or is there any other interestingness measure for rules derived from clodsed itemsets.
Regards
Re: Interestingness measure for closed itemsets
Posted by:
vinod
Date: April 23, 2014 04:18AM
If the antecedent is not a closed itemset, then after calculation of support of the antecedent, confidence can be calculated.
Re: Interestingness measure for closed itemsets
Posted by:
Philippe
Date: April 23, 2014 09:26AM
As said above, you can calculate the support of the antecedent.
There is different ways to calculate the support of the antecedent.
First, you could scan the database again to calculate the support of the antecedent but that would be costly.
If you want to find only some non-redundant rules, you could check the IGB basis of association rules. According to this basis, a rule X --> Y is created by using a generator itemset X and a closed itemset Y. There are some algorithms such as Zart that can find the generator and closed itemsets at the same time. Then this can be used to generate the IGB association rules. The code is in SPMF.
Alternatively, you could also have a look at "closed association rules". A close association rule X ==> Y is a rule where X U Y is a closed itemset. To generate the closed association rule, there is an algorithm in SPMF. You could have a look at that.
...
There is different ways to calculate the support of the antecedent.
First, you could scan the database again to calculate the support of the antecedent but that would be costly.
If you want to find only some non-redundant rules, you could check the IGB basis of association rules. According to this basis, a rule X --> Y is created by using a generator itemset X and a closed itemset Y. There are some algorithms such as Zart that can find the generator and closed itemsets at the same time. Then this can be used to generate the IGB association rules. The code is in SPMF.
Alternatively, you could also have a look at "closed association rules". A close association rule X ==> Y is a rule where X U Y is a closed itemset. To generate the closed association rule, there is an algorithm in SPMF. You could have a look at that.
...