Which algorithm is best suited for Recommendation System problem?

recommendation_engin

#1

I tried KNN algorithm and getting accuracy of about 48%, with Multi-nominal logistic regression the first iteration didn’t run properly and stopped after 100 iteration.(I created dummy variables for categorical attributes)
with SVM the algorithm is taking long time to run for even 25K records

Which algorithm you others have used for this and how much is the accuracy?


#2

Hi @sanjukta_mitra - Please go through this thread for its discussions


#3

Why not try out turicreate package by Apple.


#4

You normally tend to use something like factorization machines for recommendation problem


#5

what about the LightFM library for recommendation system


#6

try to use linear SVC algorithm. am working on it.


#7

I’ve been using the built-in Spark collaborative filtering algo for implicit feedback (i only have behavior, no ratings) and it’s working pretty well in production for a video streaming application.