Tag Boosted Hybrid Recommendation System for Multimedia Data

Chhapariya, Vinod and Singh, Manish (2018) Tag Boosted Hybrid Recommendation System for Multimedia Data. Masters thesis, Indian Institute of Technology Hyderabad.

Text
Thesis_Mtech_CS_4078.pdf - Submitted Version
Restricted to Repository staff only until July 2020.
Download (3MB) | Request a copy

Abstract

Collaborative recommendation systems are more popular for multimedia data compared to content- based recommendation system. Existing content-based recommendation algorithms give low-quality recommendations for multimedia data due to lack of good content-based features. Collaborative algorithms do well only when suficient user history is available. However, it gives very poor perfor- mance compared to content-based algorithms for new users or new items due to lack of user history. To get best of both, one can use hybrid recommendation system that integrates content and collab- orative algorithms. For multimedia data, it is di�cult to build a hybrid recommendation system as relevant content based features are not easily available. With the advent ofWeb 2.0, a lot of feedback from users on these multimedia objects is available in the form of tags, reviews, likes, comments, etc. In this paper, we propose a hybrid recommendation system that uses features from such social interactions. We do collaborative filtering using probabilistic matrix factorization and content-based filtering using topic modeling and integrate them using Bayesian model. Extensive experiments on real-world dataset show that our algorithm significantly improves the recommendation performance for multimedia data.

[error in script]

IITH Creators: