RECOMMENDER SYSTEM FOR JOB POSTING PLATFORMBack Case Studies
Client: Internal project (competition)
Length: 1.5 month
Goal: Given a new job posting, the goal is to identify those users that (a) may be interested in receiving the job posting as a push recommendation and (b) that are also appropriate candidates for the given job.
Tech: Scala, Spark
The task was defined for the RecSys 2017 competition. The detailed description can be found at: http://2017.recsyschallenge.com. The contestants were asked to make the best recommendations for the xing.com job platform. A large historical dataset of user reactions (clicks) to recommended jobs was provided for model training and testing. The goal was to make recommendations that would be the best match for both user and recruiter.
A set of relevant modeling features was created: features that describe similarity (e.g. match user vs. job industry, career level etc; calculate the job title vs. user roles similarity etc.), temporal features (e.g. capturing a user’s behaviour activities in a recent time window etc.). The dataset was labeled based on the user’s reaction, including cases when the user did not click on a recommended post. Classification + regression methods were used to distinguish and rank jobs to be recommended to users.
The evaluation metric was defined within the challenge description. Our team of 2 data scientists achieved the result that ranked us in the middle of the scoreboard of approximately 80 teams.