Analyzing Social Relations for Recommending Academic Conferences

Where should an academic publish their newest work? We developed a recommender system for it. We presented this work as a paper at the HotPOST 2016 workshop (co-located with ACM MobiHoc 2016). You can find the full paper here (PDF).

We analyze to what extent social relations from existing data can be utilized to generate academic conference recommendations. We designed and implemented a social recommender system and show how, without the need for explicit ratings, viable recommendations can be made, while at the same time reducing the cost of kNN-neighborhood selection. We evaluated our approach with data crawled from the ACM Digital Library.

The relationships considered in our system.
Graph-based modeling of the relationships.
Properties of the crawled dataset.