what are some implications/recommendations for this market basket data?

skysky Member Posts: 7 Contributor I
edited December 2018 in Help

Please provide some things you would recommend to the store manager based on this data (attached). 

1.jpg 105K

Best Answer

  • lionelderkrikorlionelderkrikor Posts: 757   Unicorn
    Solution Accepted

    Hi @sky,

     

    I was not very clear in my example. Let's detail a little : 

    A customer X buy "root vegetables" and only "root vegetables", so his current basket is "root vegetables".

    What is the closest match in Premises to his current basket : it is "root vegetables" (lines 5 & 6 ).

    Now if you choose :

     - the top matching rule in term of confidence, you will recommend "whole milk" (confidence = 0,449).

     - the top 2 matching rules in term of confidence, you will recommend "whole milk" (confidence = 0,449) and "other vegetables" (confidence = 0,435).

     

    I encourage to test some fictive baskets with RapidMiner using Apply Association Rules operator.

     

    I hope it's clearer.

     

    Regards,

     

    Lionel

     

Answers

  • lionelderkrikorlionelderkrikor Moderator, RapidMiner Certified Analyst, Member Posts: 757   Unicorn

    Hi @sky,

     

    Here my opinion on the subject : 

    In a market basket analysis, you don't provide general recommendations to the store manager, but custom recommendations to a customer X based on his current basket.

     

    The principle is the following : 

    A new customer X has a basket.

    Based on the Association Rules you builded and the current basket of customer X, you find in the Premises the closest match

    to current basket. You find top matching rules in term of hightest support / hightest confidence, and so you determine the associated Conclusions and thus the recommendations.

     

    For example, in your case, basically, if a customer buy some "root vegetables", you will recommend to him some "whole milk".

     

    In addition, take a look to the tutorial of the Apply Association Rules operator.

     

    I hope it helps,

     

    Regards,

     

    Lionel

     

     

  • skysky Member Posts: 7 Contributor I

    why did you choose "root vegetables" and  "whole milk", not "whole milk" and "other vegetables, root vegetables" (highest confidence %)? do you make a compromise between confidence and support? please help

  • skysky Member Posts: 7 Contributor I

    "You find top matching rules in term of hightest support / hightest confidence"

    how do you make that calculation to determine the best (top matching) rule?

Sign In or Register to comment.