Online Program Home
My Program

Abstract Details

Activity Number: 449
Type: Contributed
Date/Time: Tuesday, August 2, 2016 : 2:00 PM to 2:45 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #321796
Title: Predicting Job Application Success with Two-Stage, Bayesian Modeling of Features Extracted from Candidate-Role Pairs
Author(s): Jon Krohn* and Gabe Rives-Corbett and Ed Donner
Companies: untapt and untapt and untapt
Keywords: high-dimensional data ; structured features ; unstructured features ; natural language processing

We describe a two-stage model for identifying the best-suited candidates for a given role that (1) learns hiring manager preferences for the role; (2) can be updated frequently, with low latency; and (3) scales to a very large number of roles. In the first stage, dozens of candidate-role features are modeled in a regression across all applications and roles. The feature weights are subsequently fed as priors into n individual Bayesian models representing n roles. Cross-validated results indicate this approach improves classification accuracy, with the area under the curve of the receiver operating characteristic improving from 71.8% to 77.0%.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

Copyright © American Statistical Association