Re: Memory Errors in some sequential pattern mining algorihtms
Date: October 16, 2019 07:46AM
Hi,
BIDE+ is the only algorithm in SPMF that has a known bug that we did not fix yet. The problem is that BIDE is a very complex algorithm and hard to implement. I have implemented twice for SPMF, the second time starting completely from scratch, and after spending many days, it still has some issue. The bug happens when the input file contains sequences where itemsets have more than one item. If each itemset has only a single item, I think the result is correct. But if itemsets of a sequences contains more than one item, it is possible that some patterns may be missed by the algorithm.
I have thought for a while about fixing the BIDE algorithm but since there are other algorithms in SPMF that have the same input and output, and BIDE is quite complicated to debug, I think I will eventually remove BIDE from SPMF rather than fixing it. That is my plan, but I have not removed BIDE yet.... But since my implementation of BIDE has no bug for sequences of items rather than sequences of itemsets, I may perhaps keep that part of the implementation and remove the ability to apply BIDE on sequence of itemsets in the future.
In any case, I dont think that I will remove BIDE soon, so dont worry. You can still use it if you want ;-)
Best regards,
Philippe
Edited 1 time(s). Last edit at 10/16/2019 07:48AM by webmasterphilfv.