Research On Context-aware Recommendation Algorithms

Recommendation systems as a solution to information explosion have been applied in many areas such as e-commerce, social networks, and entertainment. The collaborative filtering algorithm based on looking for similar neighbors is the most successful technique in recommendation systems. But it may cause the cold-start problem when making recommendation for new users or users who have little history information because of the sparseness of the rank matrix.Furthermore, the computing mode has been changed from desktop computing to ubiquitous computing as the development of context-aware technique. The new computing mode in a ubiquitous environment provides a recommendation system with large amounts of context information including time, location, age, occupation, item property and so on. The context information has great influence on the recommendation quality. But most traditional recommendation algorithms paid little attention to the user-item-context association that limits the improvement of the recommendation quality.In our work, we first propose a recommendation approach that based on the context and a user-item relationship bipartite graph to deal with these two problems. Then we propose the spreading activation algorithm to analyze the bipartite graph. At last, our experiments show that our new approach achieves good recommendation quality solves the cold-start problem.