Improve Apriori efficiency

[rasbt]

This is an issue to revive the awesome work on the apriori algorithm in #646. This issue is partly to bump this up before the code in master further diverges from this excellent PR.

[Nachiket18]

Since frequent itemset mining is computationally intensive we use cython to speed up the code?

[rasbt]

You could do that, but I think reviving the PR linked there would be a recommended first step, and then someone can always add Cython on top of it.

[Nachiket18]

Okay that sounds good. Should we go through research papers about improvements in apriori algorithm along with the original as mentioned in the PR?

@UltraArceus3 @ankithn30 @RonakGSahu would be interested in participating.

[rasbt]

Good question (and sorry about the late response, I've been both moving and prepping a conference workshop). I would say that the improvements in the PR is the highest-impact one because the original apriori I implemented is a bit inefficient, and the "real" one from there should be much faster.