Using Hidden Markov Models to Identify Job Seekers from Social Network Data

• Activity Number: 336
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Marketing
• Author(s): Peter Ebbes*+ and Oded Netzer
• Companies: HEC Paris and Columbia University
• Keywords: Social networks ; Hidden markov models ; Marketing analytics ; Big data ; Job seeking ; Targeting

Abstract

LinkedIn is the largest professional social network in the world. One of LinkedIn's primary revenue sources comes from targeting job seekers. A key challenge for LinkedIn is to identify job seekers, as most job seekers will not publicly announce they are seeking for a job. However, job seeking behavior can be indirectly observed through how job seekers use LinkedIn. In this research we use longitudinal individual-level social network activity data from a large sample of LinkedIn members to identify job seekers. We build a hidden Markov model (HMM) in which the different states correspond to different levels of job seeking, where each state is characterized by a multivariate set of behaviors in the social network site. In addition to the longitudinal social network activity data we leverage a two-wave large-scale survey asking members directly about their job seeking status. Hence, in some time periods, and for some users, we observe their "true" job seeking status. We use the HMM to capture the dynamics in job seeking behavior as well as fuse the survey data with the longitudinal social network activity in a natural way.