A Triangle Multi-level Item-Based Collaborative Filtering Method that Improves Recommendations

Abstract

One of the most successful approaches that can provide a relevant recommendation in various domains is collaborative filtering. Although this approach has been widely applied, there are still limitation to be overcome in this research area. Accuracy is still one of the areas that needs to be improved. In addition, the rapid growth of information available online presents recommender systems with several challenges. More specifically, data sparsity and coverage affect the quality of the recommendations that can be provided. In this paper, we propose an itembased collaborative filtering (IBCF) approach with triangle similarity measures that take into account the length and angle of rating vectors between users and allow a positive and negative adjustments using a multi-level recommendation approach. We have improved the predictive accuracy and effectiveness of the proposed method, which outperforms all the compared methods in terms of the mean absolute error (MAE) and the root mean squared error(RMSE). We aimed to evaluate the proposed method by comparing our results with those of some popular similarity mesures using k-nearest neighbour (kNN) algorithms. We ran our experiment using three real dataset: MovieLens 100K, MovieLens 1 M and Yahoo! Movies.

A novel recommendation method based on general matrix factorization and artificial neural networks

“…The evaluation results show great efficiency for the proposed method, however compared with previous work [1], Multi-level CF and PCC still show to be superior by a small percentage. Experimentation with GMF and ANNs shows a competent and fast approach to converge in recommender datasets, however more advanced neural network similarity functions should be experimented to extend the experimental coverage of possible efficient advanced recommender system techniques.…”

“…Thus, modified similarity measures is one of the most important research challenges in recommender systems, because the prediction accuracy in recommender systems can be improved substantially (or not) according to the applied method. In this paper we extend previous work [1] by introducing a recommendation method that is based on general matrix factorization and artificial neural networks. The main contributions of this paper are as follows:…”

“…3 The Proposed Method Previous work [1] has proposed a multi-level triangle recommendation method, which is explained in section 3.1 below, that improved the recommendation accuracy and effectiveness outer performing all baselines using MAE. To automate further this approach, in this paper we propose the use of a method based on general matrix factorization (GMF) and artificial neural networks (ANN) which is explained in section 3.2.…”

A switching multi-level method for the long tail recommendation problem

“…Recently, many authors presented a modification to the similarity function to improve the CF recommendation [12,13,14,15]. For example, a multi-level method was proposed by the authors in [12] that utilize a nbumber of constraints that enhance the Pearson correlation coefficient (PCC) similarity value of users who belonged to specified categories based on the number of corated items and the minimun PCC value between them.…”

“…This methdod is explained in more detail towards the end of this section. In addition, in [15], another multilevel approach was presented which is similar to the aforementioned but the triangle similarity was used, which improves the prediction accuracy even more. In [14], the cosine similarity was modified using co-rated items as an adjusted factor to improve the similarity.…”