HYBRID WHALE OPTIMIZATION ALGORITHM — K-MEANS CLUSTERINGFOR RECOMMENDER SYSTEM

Abstract

Recommender systems apply various techniques to come up with the best suggestions that would match an ideal customer preference. The systems are founded on filtering approaches that are either content-based filtering, collaborative filteringor hybrid filtering which is a combination of both. These filtering techniques employ various machine learning algorithms.In this paper, the Whale Optimization Algorithm with K-means (WOA-K-means) clustering is employed to determine its efficiency in recommender systems. The data used is the MovieLens 100k dataset that contains the movie reviews. The data is used in developing the recommender system as well as testing its efficiency based on the ratings that determine the customer preferences. The obtained results by the proposed WOA-K-means algorithm are compared withtraditional K-means clustering algorithm as well as other powerful algorithms like Principal Component Analysis K-means (PCA K-means), Self-Organizing Map (SOM) Cluster, PCA-SOM. It is found that WOA-K-means algorithm hassuperior precision and recall values as compared to the tested standard algorithms.WOA-K-means also has the lowest computational time of the hybrid systems. The WOA-K-means algorithm achieved a MAE of 0.6438 which is superior as compared to the existing algorithms with an average MAE of 0.75. Thus, WOA-K-means algorithm is demonstrated as a superior algorithm for developing recommender systems due to its high efficiency from the various measures. The proposed WOA-K-means algorithm will be helpful in other similar recommendation tasks.