Exploiting Graph Similarities with Clustering to Improve Long Tail Itens Recommendations
Diogo Vinícius de Sousa Silva, Amanda Chagas de Oliveira, Francisleide Almeida, Frederico Araújo Dur�ão · 2020
Techniques in recommendation systems generally focuses on recommending the most important items for a user. The purpose of this work is to generate recommendations focusing on long tail items, leading users to less popular and at the same time highly relevant products. Two techniques from the literature were applied in this study. The first technique is through graphs to calculate node similarity between users and items. The second technique applies clustering in the set of items in a dataset. This combination was adopted in order to give more visibility to long tail items. To evaluate the proposed approach an experiment was carried out to calculate the accuracy, diversity, and popularity of the generated recommendation. We compare the proposed approach with other 3 baselines where our approach achieved better results.