Association rule mining (ARM)

dongen_mdongen_m Member Posts: 3 Contributor I
edited November 2018 in Help

All,

 

I am new in RapidMiner. I need to do an association rule mining project. I have some questions:

 

1. How do I do the data cleaning before association rule mining? Do you need to clean missing values? Do you need to balance the classes? Do you need to remove outliers? etc...?

2. After the data cleaning I need to do W-Apriori or FP-Growth? When to do W-Apriori and when FP-Growth?

3. After this I need to apply a classification algorithm of your choice or/and a clustering algorithm? Do you use the best associations from W-Apriori or FP-Growth as selected attributes as input for the classification or clustering algorithm?

 

Many thanks for your help!!

 

Tagged:

Answers

  • lionelderkrikorlionelderkrikor Moderator, RapidMiner Certified Analyst, Member Posts: 716   Unicorn

    Hi @dongen_m,

     

    You can find first response elements about FP-Growth in the process template Market Basket Analysis

    GP_Growth.png

    I hope it will be helpful to begin your project.

     

    Regards, 

     

    Lionel

     

  • lionelderkrikorlionelderkrikor Moderator, RapidMiner Certified Analyst, Member Posts: 716   Unicorn

    Hi again @dongen_m,

     

    Second response elements in this ressource (see attached file) which contains : 

     - a description of Apriori algorithm

     - a description of FP-Growth algorithm

     - a detailed implementation of the FP-Growth in RapidMiner

     

    I hope it helps,

     

    Regards, 

     

    Lionel

  • balaji_sundarambalaji_sundaram RapidMiner Certified Analyst, Member Posts: 8 Contributor II

    Please confirm that the dependent variable contains Binary values

  • lionelderkrikorlionelderkrikor Moderator, RapidMiner Certified Analyst, Member Posts: 716   Unicorn

    @balaji_sundaram,

     

    It's the different attributes (for example the differents products we are looking for associations) which have to be on a Binominal format (false or true).

     

    Regards, 

     

    Lionel

Sign In or Register to comment.