An Improved Collaborative Filtering Algorithm Using Support Values

Author(s): Perez Jose Laiju, Rachel Philo Jojo, Resi T Reji, Vimal K, Mrs. Priya K V

Abstract: Internet has a large number of data which are widely diverse and huge. The Recommendation System is used to filter out these contents according to the preference of intended users. It is an information filtering system that helps in prediction or rating, based on user preference. Conventionally, collaborative filtering is the common approach to the design of recommendation system. This approach uses the similarity measures which rely on the similarities between users. The method that we use here is a combination of multiple similarity measures, that are used to find similarities between users. Some of the similarity measures that we have used here are Pearson Correlation Coefficient, Mean Square Difference, Cosine Similarity, etc. Using the method of aggregation, we divide it into super similar and super dissimilar users and different similarity values are assigned to each. We use Root Mean Square Error to evaluate the predictive accuracy.