APPLICANT RECIPROCAL RECOMMENDER

THE CHALLENGE

Our client, a referral based recruitment portal, wanted to improve their ranking algorithm for candidates applying for a job. The current ranking algorithm was based on matching candidate skills with the job description. These were found to be sub optimal.

OUR APPROACH

In order to arrive at better rank for a candidate for a job application, merely matching the candidate skills and job description is not enough. There also needs to be a cultural match of the candidate and the quality of referrer should also be factored in.

THE METHODOLOGY

  • Develop spiders to collect data from multiple sites on the Internet to
    • quantify the quality of candidate skills
    • characterize companies and educational institutes to predict the cultural match between the candidates employment history and prospective employer
  • Extract information from spidered text to develop an influence network to measure the “network worth” of a referrer using social network analysis.
  • Use Genetic Algorithms to optimize a hybrid model using skills, cultural match and referrer quality

TECHNIQUES, TECHNOLOGIES, TOOLS

Text Analysis, Topic Models, Genetic Algorithm, Conditional Random Fields, R, jsoup

RESULT/IMPACT

There was a significant reduction in the number of candidates that needed to be interviewed as compared to prior to the use of the applicant recommender hence, saving a lot of time and money for the client.