k-Surrounding Neighbors: Incorporating Serendipity in Collaborative Recommendations
Huidi Lu · Atlantis Highlights in Intelligent Systems/Atlantis highlights in intelligent systems · 2023
Social recommender systems have become ubiquitous in our online environment, but there are growing concerns that they narrow our horizons and polarize our opinions.This paper proposes a new recommendation algorithm, k-Surrounding Neighbors, based on the theory of weak ties, to increase the diversity and novelty of recommendations.The proposed method discards some nearest neighbors based on their similarity, giving more weight to less similar others, which can provide fresh information and new experiences.Validation tests using several metrics show that it significantly improves recommendation diversity and novelty at a minor cost in precision.