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
Regarding class association rule mining
Posted by:
Vanitha
Date: January 20, 2015 04:14AM
Hello,
I am doing my project on class association rule mining. Your software is helping me
a lot.
Thank you very much. I want to know how to derive class association rules from
association rules. could you please explain me how to do it.
Thanks again.
I am doing my project on class association rule mining. Your software is helping me
a lot.
Thank you very much. I want to know how to derive class association rules from
association rules. could you please explain me how to do it.
Thanks again.
Re: Regarding class association rule mining
Posted by:
webmasterphilfv
Date: January 22, 2015 08:09AM
Hi,
Glad that the software is helpful.
I have read about class association rules a while ago. But basically it is rules of the form A1,A2,A3 --> Bx where the antecedent may have multiple items and the consequent has only 1 item that is a value for the target attribute.
You could first mine the association rules and then filter them to keep only the class association rules. It would work but it would be inefficient because it would need to generate many rules that are not class association rules and then to delete them. So, the best option is to use a custom algorithm that mines the class associatino rules directly.
There is no such algorithm in SPMF. But you could probably adapt some for that problem. But in my opiinion should first look at papers about class association rules such as CBA, CPAR, CMAR and other similar algorithms. They will explain how they mine the rules.
Best,
Glad that the software is helpful.
I have read about class association rules a while ago. But basically it is rules of the form A1,A2,A3 --> Bx where the antecedent may have multiple items and the consequent has only 1 item that is a value for the target attribute.
You could first mine the association rules and then filter them to keep only the class association rules. It would work but it would be inefficient because it would need to generate many rules that are not class association rules and then to delete them. So, the best option is to use a custom algorithm that mines the class associatino rules directly.
There is no such algorithm in SPMF. But you could probably adapt some for that problem. But in my opiinion should first look at papers about class association rules such as CBA, CPAR, CMAR and other similar algorithms. They will explain how they mine the rules.
Best,
Re: Regarding class association rule mining
Posted by:
Dang Nguyen
Date: January 22, 2015 05:05PM
If you primarily consider the mining time, you could study algorithms which directly generate class association rules (CARs) from the dataset. Some recent algorithms are:
1. CAR-Miner: http://www.sciencedirect.com/science/article/pii/S0957417412011591
2. CAR-Miner-Diff: http://www.sciencedirect.com/science/article/pii/S095741741500007X
3. PMCAR (Parallel CAR Mining): http://www.sciencedirect.com/science/article/pii/S0957417414000621
CBA, CMAR, and CPAR are state-of-art algorithms for CAR mining. They, however, mainly focus on the classification accuracy.
Another topic related to CAR mining is to mine CARs with the itemset constraint (i.e. the left hand side of the rule contains at least one of itemsets specified by end users). You can take a look this paper:
CCAR: http://www.sciencedirect.com/science/article/pii/S0952197614002127
Hope it helps.
1. CAR-Miner: http://www.sciencedirect.com/science/article/pii/S0957417412011591
2. CAR-Miner-Diff: http://www.sciencedirect.com/science/article/pii/S095741741500007X
3. PMCAR (Parallel CAR Mining): http://www.sciencedirect.com/science/article/pii/S0957417414000621
CBA, CMAR, and CPAR are state-of-art algorithms for CAR mining. They, however, mainly focus on the classification accuracy.
Another topic related to CAR mining is to mine CARs with the itemset constraint (i.e. the left hand side of the rule contains at least one of itemsets specified by end users). You can take a look this paper:
CCAR: http://www.sciencedirect.com/science/article/pii/S0952197614002127
Hope it helps.