On the Impact of Neighborhood Selection Strategies for Recommender Systems in LBSNs

Selecting and weighting users in collaborative filtering based POI recommendation

“…In this paper, we used a strategy for selecting neighbors based on graph of common visited places. In a previous work [30], this strategy outperformed other strategies considered for selecting neighbors.…”

“…We have combined these strategies with a neighbor selection approach that is based on a graph of common visited places. This strategy outperformed others in the experimental evaluation reported in a previous work [30]. We carried out different experiments to evaluate the strategies proposed against a traditional user based collaborative filtering approach.…”

“…In contrast to the described works, this paper aims to assess the impact of the different strategies proposed in the scoring of neighbors in user-based CF-based POI recommendation. In a previous work [30], strategies for selecting neighbors were studied, showing the advantage of building a graph of common visited places. The scoring strategies evaluated considered the different elements available in LBSNs, such as places categories and social information.…”

“…As userbased CF trusts neighbors as information sources, the quality of recommendations is a direct consequence of the selected neighborhood and the importance given to each neighbor for prediction. In a previous work [30], we have studied different strategies for selecting neighbors considering the information contained in the LBSNs. We have concluded that selecting as neighbors those users that have visited the same places as the target user we can obtain lower errors in the preference estimation process.…”