Diversifying the predictions in the recommender systems

In pursuance of building a recommender system, the existing collaborative filtering model often fails to provide a diversified list of recommendation to the end user. Most of the existing models target accuracy and thus fails in avoiding the uniformity dullness from the recommended list. In our paper, we have made use of imputation technique to cut-off sparsity and employed graph-based algorithm to generate a diversified list of recommendations in order to prevent the aforementioned problem of over specialisation. A substantial coverage evaluation on MovieLens dataset demonstrates the fruitfulness of our proposed graph-based model.