The recently launched LinkedIn Salary product has been designed with the goalof providing compensation insights to the world's professionals and therebyhelping them optimize their earning potential. We describe the overall designand architecture of the statistical modeling system underlying this product. Wefocus on the unique data mining challenges while designing and implementing thesystem, and describe the modeling components such as Bayesian hierarchicalsmoothing that help to compute and present robust compensation insights tousers. We report on extensive evaluation with nearly one year of de-identifiedcompensation data collected from over one million LinkedIn users, therebydemonstrating the efficacy of the statistical models. We also highlight thelessons learned through the deployment of our system at LinkedIn.
展开▼
