Instituteof Mechanical Electronic and Information Engineering, China University of Mining and Technology Beijing, Beijing, China.

---

Abstract—Most of the popular recommendation algorithms are providing similar recommendations to users based on their ratings. However, in terms of competition, past records of user ratings do not directly translate to users’ interests in new competitions. Competitions are challenges, and as such, users are unlikely to choose what they have registered before, but instead, prefer challenges that complement the scope of their present abilities. Thus measurements of a user’s interest in competitions should be based on the differences, rather than similarities, in user’s past registration data. In this paper, we propose an alternative recommendation algorithm that measures users’ interests in competitions based on these differences.First, competition differences, such asregistrations, stars and browsers records are modeled and calculated. Then, the peak values and the range of users’ interests are attained through such differences. Finally, recommendations of competitions are made if they fall within the range radius. The proposed algorithm proves to be more effective and efficient than conventional recommendation algorithms due to its consideration of competition’s features as well as the user’s psychology.

Index Terms—Competition recommendation method, disparity measurement on participation interest, disparity

[PDF]

Cite: Kaixin Liu,Jianhui Chang, Huiwen Ren, Hong Zhu, "Disparity-Based Measurement on Participation Interest and Competition Recommendation Method," Journal of Software vol. 13, no. 2, pp. 117-125, 2018.