A Fair Group Recommendation System Based on Members and Leader Influences
In a group recommender system, the effort is made to provide recommendations to a group of individuals rather than a single person. In these systems, the opinions of all group members are influential in decision-making, aiming to provide the best choice despite different personal preferences. This a...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | fas |
| Published: |
Semnan University
2024-08-01
|
| Series: | مجله مدل سازی در مهندسی |
| Subjects: | |
| Online Access: | https://modelling.semnan.ac.ir/article_8376_9c26313da750548490e9bd5d9a52c039.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In a group recommender system, the effort is made to provide recommendations to a group of individuals rather than a single person. In these systems, the opinions of all group members are influential in decision-making, aiming to provide the best choice despite different personal preferences. This article attempts to present a group recommender system capable of identifying the relationship among users and eventually determining the influence of each user on the group, subsequently offering the best recommendations based on these connections. Moreover, a new criterion for determining leadership in the group is introduced, which identifies the leader of the group based on the level of trust, similarity, belongingness, and dependence of users on the group. Additionally, a novel criterion for delivering fair recommendations to the group is proposed, suggesting items to users with the most positive feedback among all group members. The proposed algorithm is compared with similar algorithms in this domain in two sections. In the evaluation section of assigned rankings, the accuracy of the proposed method was close to 100% in all cases, reporting an average improvement of 5% compared to the compared methods. In the recommendation evaluation section, well-known criteria such as nDCG, group satisfaction, and fairness were used, where the proposed method showed an average improvement of 41%, 35%, and 38%, respectively, considering the number of diverse recommendations in each of the mentioned criteria on average. |
|---|---|
| ISSN: | 2008-4854 2783-2538 |