CROSS DOMAIN RECOMMENDATION USING VECTOR MODELING AND GENRE CORRELATIONS

Abstract

Recommender systems are basically information retrieval systems that offer guidance to users in making individual decisions related to choosing items based on personal interests. On Internet, there are infinite numbers of results for a particular query like movies, music, books, clothes etc. Sorting through every result is very tedious and time-consuming. Recommender system is very important application of data science and machine learning. They make the job of recommendation and prediction of preferences of users very simple. There are many limitations in classical recommender system because they provide recommendations in single domain only. With proliferating e-commerce sites and limitations in collaborative and content based recommender systems, cross domain recommender system are now widely in use. They can address the data sparsity and cold start problem by utilizing data from other related domains. In this paper, we propose recommendations across different domains by combining the benefit of plot keywords extracted from storyline and genre details from the two entertainment domains. We illustrate the working of our proposed CDR scheme using the movie as source domain and book as target domain

Keywords

• Cross Domain Recommender System
• Cold Start Problem
• Keywords
• Vector Modeling
• Genre Correlation