PREFERENCE WEIGHTED RESTRICTED BOLTZMANN MACHINES FOR COLD START RECOMMENDATIONS

Abstract
Collaborative filtering and content filtering are the two techniques used in filtering data for Recommender Systems. One of the not fully resolved problems of Collaborative filtering is the cold start problem and is still an active research topic in the area of Recommender Systems. The cold-start problem of items is still a great research topic to work on. With the advent of Machine Learning there seems to be some success though; different models and algorithms with diversified approaches are emerging. In this work, we make use of Boltzmann machines and specifically a Restricted Boltzmann machine (RBM) with dynamic data to solve the cold start problem. We propose a collaborative filtered model with RBM and call it CF_RBM. The active and passive browsing history and the time spent on a website are taken as dynamic data and passed along with the other visible units of the RBM. The user preferences of nearly 100 random users of the system was also included. We consider the Movielens 10M dataset and apply the CF_RBM model. The recommendation results obtained along with the recommendation score shows that the new items are generated, hence resolving the cold-start problem.

Authors
G.S. Ananth1, K. Raghuveer2, S. Vasanth Kumar3
The National Institute of Engineering, IndiaIndia1, The National Institute of Engineering, India2,Mangalore Institute of Technology and Engineering, India3

Keywords
Recommender Systems, Intelligent Systems, Decision Support Systems,Collaborative Filtering, Boltzmann Machines, User Preferences,Passive Browsing History
Published By :
ICTACT
Published In :
ICTACT Journal on Data Science and Machine Learning
( Volume: 3 , Issue: 3 )
Date of Publication :
June 2022

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.