Longest Common Subsequence based Multistage Collaborative Filtering for Recommender Systems

The contemporary recommender systems are facing challenges such as noise in user's choice, scalability, cold-start problem, availability of ample choices and handling of sparse data sets. In this paper, a multistage collaborative filtering is proposed to address the issues of noise in user's choice and ample choices availability. The two-stage filtering at first stage, filtering is performed using Pearson coefficient as a similarity measure and in the second stage, the longest common subsequence (LCS) is used to do filtering. The experiments are performed using benchmark 100k movielense datasets. The performance of multistage collaborative filtering is evaluated using accuracy, precision, recall, and f-measure. The results are also compared with single stage filtering and performance of multistage collaborative filtering is significantly improved over the used datasets.