I'm using spark mllib for FP-Growth algorithm for our ML model.
Description of my issue:
I have taken transactional data from our production database to mine the frequent brought items recommendation. Now when I run the data with this algo I always get recommendations on the products that are sold the most. For ex. For a t-shirt I get recommendations of t-shits itself instead of something like pants etc..
Now one thing which I have Noticed is the item t-shirt in above example is present in dataset for 65k times and pant is appearing only in 2k cases.
how do I now optimise the data to get a better prediction?