A ranker ensemble for multi-objective job recommendation in an item cold start setting
11th ACM Conference on Recommender Systems Challenge Workshop, RecSys Challenge 2017, Como, İtalya, 27 Ağustos 2017, cilt.Part F130525, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: Part F130525
- Doi Numarası: 10.1145/3124791.3124798
- Basıldığı Şehir: Como
- Basıldığı Ülke: İtalya
- Anahtar Kelimeler: Cold start, Implicit feedback, Multi-objective optimization, Recommender systems
- Boğaziçi Üniversitesi Adresli: Evet
Özet
Real-life recommender systems often havemultiple objectives, and considering only a single stakeholder's perspective can be inadequate. ACM Recsys 2017 challenge requires such a multi-objective job recommender system taking into account job seeker satisfaction, recruiter satisfaction, balance between relevance and revenue, and scalability in a primarily item cold start setting. In this paper, we discuss our findings, and a possible solution to this problem. Making use of interaction profiles and content data, a hybrid of ranking algorithms is proposed to optimize the required objectives. This solution was able to achieve 8th position during A/B testing, and some of the ideas can be useful in a more complex ensemble.