Abstract
Recommendation system helps filter out relevant and concise information from large internet data in least possible steps and time. Recent research on recommendation techniques steadily evolved in finding possibilities of improvement and exploration of novel techniques for better recommendation approach. Graph based hybrid techniques attained significant attention in recent time to improve recommendations. While collaborative filtering and other state of the art techniques perform well, this work proposes hybrid graph based approach utilizing connections in heterogeneous information graph, integrating significant relation among features to recommend fresh and rich recommendations. In this work, we propose weighted meta-paths in relationship graph where we compute relationship strength through feature aggregated score to recommend. We test our proposed methods on movie datasets and show its relevance to the users through recommendations. Insisting on positive and relevant features to user's profile, we recommend results based on proposed meta-paths and its features. Adding features does not always guarantee relevance of recommendation, and to explore more, we validate recommendations through tests to evaluate its novelty and relevance in comparison with other state of the art methods..