A Hybrid Collaborative Filtering Approach for Recommendations

Collaborative recommender system has been an important and popular approach in making recommendations. However, it suffers with the cold start and sparsity problems. Therefore, to alleviate the problems, we have combined a set similarity and user evaluation method in collaborative filtering by introducing some additional weight parameters. Further, we optimize these parameters by Particle Swarm Optimization (PSO). We named it as hybrid approach. The hybrid approach is tested on movie recommendation to the user. The experimental result confirms that the mean absolute error with the other approaches is encouraging. Keywords— Set Similarity, Recommendation, Cosine Similarity, Prediction, Collaborative Filtering, Ontology.