Sequence Mining

[scikit-mine]

Sequence Mining

[remiadon]

https://github.com/chuanconggao/PrefixSpan-py

This repo implements PrefixSpan, BIDE and FEAT

Amongst those 3 BIDE seems the most interesting, as it:

• discovers closed sequences (consistency with LCM, which discovers closed itemsets)
• has the most citations counts (826)
• can be implemented efficiently (?)

Those closed sequences can be considered as candidates for SQS-candidates, which starts from a pre-mined set of sequences.
Note that FEAT discovers the set of generator sequences, which seems to be closer to an MDL optimal representation thant closed sequences. So perhaps using generator sequences as candidates to SQS-candidates would be more efficient ...