The Data Mining Forum                             open-source data mining software data science journal data mining conferences machine learning in software engineering MLISE 2021 utility mining workshop at ICDM 2021
This forum is about data mining, data science and big data: algorithms, source code, datasets, implementations, optimizations, etc. You are welcome to post call for papers, data mining job ads, link to source code of data mining algorithms or anything else related to data mining. The forum is hosted by P. Fournier-Viger. No registration is required to use this forum!.  
Question regarding FHM+
Posted by: Carter
Date: July 28, 2021 01:52PM

Hello,

I work in the ecommerce department for a semi-large organization. I have researched SPMF and all of its individual algorithms very much.

My company only wants pairs of relationships as the output. I first started by using FPGrowth_association_rules to get these pairs. I set the Max antecedent and Max consequent length to 1 in order to return only pairs. I also set the min support and min confidence very low as well.

After getting these results I went to use FHM+ to implement the utility in. The thing is I have to use the same input I used for the FPGrowth (while adding their utilities) but then I am not able to choose and Max antecedent or Max consequent.

At first I thought this was the same as the min/max length parameter that FHM+ has but I realized this is only choosing invoices that have a maximum length to what I set to.

Conceptually, is there any way I can achieve this goal? I cannot use the pairs from the FPGrowth since they are scrambled and do not carry their individual utilities with them.

Thank you for any insight!

- Carter

Options: ReplyQuote
Re: Question regarding FHM+
Date: August 13, 2021 08:31AM

Dear Carter,

I am happy to hear about your tests using the SPMF software. I am sorry to answer almost two weeks later. It is here the vacation time in China and I did not check the forum for a while.

It is indeed a good idea to set these maximum constraints to filter longer patterns.

While FPGrowth can be used to generate rules, the FHM+ algorithm is only designed to generate itemsets (sets of values) rather than rules like A --> C. This is the reason why there is no max antecedent and max concsequent. The concept of antecedent and consequent only exist for rules. The antecedent is the left side of the rule and the consequent is the right side.

FHM+ will find itemsets like {a,b} indicating that a and b appear together and have a high utility. But this is different from a rule like a--> b or b --> a.

If you want to generate association rules with utility, there is no algorithm for this in SPMF. If you really need this, maybe it could be done in another way? or some algorithm could be adapted? I am not sure... Maybe it depends on your application.

Again, I answer you quite late. Hope you have found some other ideas to solve the problem you were facing!

Best regards,

Philippe

Options: ReplyQuote


This forum is powered by Phorum and provided by P. Fournier-Viger (© 2012).
Terms of use.