Best item recommendation system algorithm for production level

r
rmachinelearning
machine_learning
datascience

#1

Hello guys currently i’m using association rule for item based suggestion. Please tell me it is good or i should move on IBCF algorithm or which algorithm is best for item recommendation
Kindly suggest me


#2

Hey Sushil,

I think the question you asked is very broad in its sense. A recommendation engine could serve different use-cases that occur at different carousels or pages on a web portal. What is the aim of your recommendations? Are you trying to increase the AOV? Are you trying to reduce friction? Or maybe upsell?

To give a generic answer - you association rule mining would not give you enough rules. So let’s say you want to recommend 5-10 products for a given product (also depends on your catalog size and corresponding purchase history), you may not find enough items to recommend with the rules. Also, if your catalog has new items getting added, it will not include those since it won’t meet the required thresholds for confidence/lift that you set.

There could be different ways to go about it depending on what you are trying to achieve.
Hope this helps.


#3

Actually i want to create customer personas and my problem is how i predict customer favourites items combination like he take starter with liquor or prefer beer with main course.
So how we create that personas at customer level in machine learning.