Association Rule Learner (Borgelt) find optimal minimum support value


I’m trying to run the Association Rule Learner (Borgelt) for different data sets (in a loop). For each dataset I’ve determined the minimum support which I control using a flow variable. I want to run the node each month. Every month the input data changes slightly. Some months I experience that the minimum support from last month is too low which causes the Association Rule Learner (Borgelt) to get stuck in “process running for X seconds”. Sometimes it gets up to 1000 seconds and when it finally does goes to the next phase of parsing this also takes a very long time, because there are obviously a lot of rules.

I was wondering if there is some way to run a process or a model or something where each month I can find the ideal minimum support automatically so that the process doesn’t run for too long but also doesn’t give an empty table (when minimum support is too high). I don’t want to keep adjusting the minimum support manually.

When I was thinking towards a solution I thought maybe there is a way to cancel a node when it’s running too long (for example more than 200 seconds) and then increase the minimum support with 1% or something and repeat this process until the node runs smoothly. But I haven’t found a way to actually implement something like this.

Can anyone help me with finding the ideal minimum support in an automated process?

Winanda Sheombarsing


Hi @Winanda,

I am afraid that it is not possible to cancel a running node automatically from within a workflow. Only option is to cancel it manually. So therefore I also don’t see any workaround, although I like your idea with the loop that sequentially increases the support parameter after a time out.
I will file a feature request, but obviously no promises there. Thanks anyway for your input! Appreciated.



This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.