Recommendation system using the k-nearest neighbors and singular value decomposition algorithms

Abstract

<span>Nowadays, recommendation systems are used successfully to provide items (example: movies, music, books, news, images) tailored to user preferences. Amongst the approaches existing to recommend adequate content, we use the collaborative filtering approach of finding the information that satisfies the user by using the reviews of other users. These reviews are stored in matrices that their sizes increase exponentially to predict whether an item is relevant or not. The evaluation shows that these systems provide unsatisfactory recommendations because of what we call the cold start factor. Our objective is to apply a hybrid approach to improve the quality of our recommendation system. The benefit of this approach is the fact that it does not require a new algorithm for calculating the predictions. We are going to apply two algorithms: k-nearest neighbours (KNN) and the matrix factorization algorithm of collaborative filtering which are based on the method of (singular-value-decomposition). Our combined model has a very high precision and the experiments show that our method can achieve better results.</span>

A hybrid recommender system based on customer behavior and transaction data using generalized sequential pattern algorithm

“…Recommender systems offer products or services according to the users' preferences [25] by utilizing common data such as ratings, reviews, and feedback [26]- [28] to generate personalized recommendations [29], [30]. Recommender systems can be classified into several types based on the data used to generate recommendations.…”