Publication

Abstract : Recommender system (RS) is an emerging technique in information retrieval to handle a large amount of online data effectively. It provides recommendation to the online user in order to achieve their correct decisions on items/services quickly and easily. Collaborative filtering (CF) is one of the key approaches for RS that generates recommendation to the online user based on the rating similarity with other users. Unsupervised clustering is a class of model-based CF, which is more preferable because it provides the simple and effective recommendation. This class of CF suffers by higher error rate and takes more iterations for convergence. This study proposes a modified fuzzy c-means clustering approach to eliminate these issues. A novel modified cuckoo search (MCS) algorithm is proposed to optimize the data points in each cluster that provides an effective recommendation. The performance of proposed RS is measured by conducting experimental analysis on benchmark MovieLens dataset. To show the effectiveness of proposed MCS algorithm, the results are compared with popular optimization algorithms, namely particle swarm optimization and cuckoo search, using benchmark optimization functions.