Affiliations: [a] Computer Science and Engineering, National Institute of Technology Nagaland, Dimapur, India
| [b] Computer Science and Engineering, Indian Institute of Information Technology Tiruchirappalli, Tiruchirappalli, India
Correspondence:
[*]
Corresponding author. Bam Bahadur Sinha, Computer Science and Engineering, National Institute of Technology Nagaland, Dimapur, India. E-mail: [email protected].
Abstract: In the current era of big data, the recommender system aspires to provide users with a tailored set of personalized items from a pool of a large population. The most popular collaborative filtering system performs this information filtering process by computing similarity among users or items. This paper proposes a similarity metric that comprises of weights and values. Values are calculated by considering the matching set of users for which similarity is to be computed. The optimal values of weights are decided using an upgraded form of the Crow Search Algorithm (CSA). The exploration and exploitation stability of CSA is improvised by making use of Levy flight diffusion, adaptive operator adjustment, and event factor. The performance of the implemented metaheuristic approach is validated on Jester, MovieLens 100K, and MovieLens 1M dataset. Comparative analysis of proposed model against several other traditional metaheuristic based personalization systems reveal that our model is less delicate to the dimension of datasets and it also presents exceptional refinement in terms of prediction complexity and accuracy.
