Clustering-Based Collaborative Recommender System Using a Nature-Inspired Algorithm

The recommender system is a filtering technique that tries to reduce the available selections for users by finding the relevant items that satisfy their desires. Collaborative filtering is the most simple and efficient recommender system approach, as it makes use of the similarity among users’ preferences to generate recommendations. However, this approach suffers from some issues such as data sparsity and scalability, which make the process of identifying user’s neighbourhood a challenging task. This paper proposes a clustering-based collaborative filtering recommender system using the k-means algorithm along with a nature-inspired algorithm to improve the clustering process, which leads to a better recommendation quality. The Artificial Fish Swarm Algorithm (AFSA) has been employed to determine the optimal initial centroid for the clustering task. MovieLens 100K dataset was used to evaluate the proposed system’s performance in terms of precision, recall, and mean absolute error. The dataset was divided into 80% and 20% randomly for training and testing, respectively. Python 3.9.12 programming languages was used to implement the proposed system. The experimental results confirmed over the different number of clusters and the value of mean absolute error decreased from 0.75 to 0.67 as the number of clusters increased from 5 to 35 clusters. The obtained results are compared against other recent nature-inspired algorithms, which proved that the proposed recommender system outperforms the considered methods.