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
Frequent Itemsets Mining
Posted by:
Pablo
Date: April 06, 2014 11:45AM
Hello everyone, I'm trying to solve the problem of given a 0/1 matrix find every submatrices of 1's having at least k rows.
In this paper
http://jmlr.org/papers/volume9/sun08a/sun08a.pdf
appears that my problem has already been solved as the Frequent Itemsets Mining (FIM) problem.
In this form the FIM problem can be stated as follows: given X and k >= 1, find every submatrix of 1s in X having at least k rows, and report the associated set of columns. If the threshold k is allowed to vary, then FIM algorithms essentially seek to find every maximal submatrix of 1s in the data matrix X.
Please is there any algorithm that solves this problem?
In this paper
http://jmlr.org/papers/volume9/sun08a/sun08a.pdf
appears that my problem has already been solved as the Frequent Itemsets Mining (FIM) problem.
In this form the FIM problem can be stated as follows: given X and k >= 1, find every submatrix of 1s in X having at least k rows, and report the associated set of columns. If the threshold k is allowed to vary, then FIM algorithms essentially seek to find every maximal submatrix of 1s in the data matrix X.
Please is there any algorithm that solves this problem?
Re: Frequent Itemsets Mining
Posted by:
khairy
Date: April 15, 2014 12:39AM
Hi Sir
It seems that in all datasets H-mine algorithm record the worst execution time compared to FP-Growth, Apriori and Ecalt. Is this true?
Thanks
It seems that in all datasets H-mine algorithm record the worst execution time compared to FP-Growth, Apriori and Ecalt. Is this true?
Thanks
Re: Frequent Itemsets Mining
Posted by:
webmasterphilfv
Date: April 15, 2014 04:00AM
Yes... But it is most likely because the implementation of HMine in SPMF is not good. I should re-implement HMine.
Re: Frequent Itemsets Mining
Posted by:
khairy
Date: May 16, 2014 10:01AM
Hi Sir,
When i try to run FP-Growth in 0.9 support using Pumsb dataset, i received this message
Exception in thread "main" java.lang.OutOfMemoryError: Java heap space
as shown i give a big support value but the algorithm is crash
how can i run the algorithm in this dataset to get frequent item-sets.
When i try to run FP-Growth in 0.9 support using Pumsb dataset, i received this message
Exception in thread "main" java.lang.OutOfMemoryError: Java heap space
as shown i give a big support value but the algorithm is crash
how can i run the algorithm in this dataset to get frequent item-sets.
Re: Frequent Itemsets Mining
Posted by:
webmasterphilfv
Date: May 16, 2014 02:22PM
I think that you should download the latest version of SPMF. There is an optimization of FPGrowth that was added a few weeks ago.
On my computer with psumb dataset and minsup = 0.9 :
============= FP-GROWTH - STATS =============
Transactions count from database : 49046
Max memory usage: 8.296775817871094 mb
Frequent itemsets count : 2607
Total time ~ 6084 ms
===================================================
4404 #SUP: 45106
4404 4428 #SUP: 44325
4404 4428 4438 #SUP: 44229
On my computer with psumb dataset and minsup = 0.9 :
============= FP-GROWTH - STATS =============
Transactions count from database : 49046
Max memory usage: 8.296775817871094 mb
Frequent itemsets count : 2607
Total time ~ 6084 ms
===================================================
4404 #SUP: 45106
4404 4428 #SUP: 44325
4404 4428 4438 #SUP: 44229