Understanding Job market using Probabilistic Graphical Models

3,629 views
3,762 views

Published on

Published in: Technology, Business
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
3,629
On SlideShare
0
From Embeds
0
Number of Embeds
2,159
Actions
Shares
0
Downloads
0
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

Understanding Job market using Probabilistic Graphical Models

  1. 1. Understanding the Job market using Probabilistic Graphical Models20 June 2012 Venkatesh Umaashankar
  2. 2. Job Market  Hiring managers go out to a job market thinking they want to hire Talent ABC; the reality is that often what theyll end up hiring is ABX, AXY, or even XYZ. The job market informs, and can shift what the hiring manager ends up hiring. what the job spec initially says is not what the candidate ends up possessing.  Source: http://www.quora.com/Hiring/Does-the-hidden-job-market-actually-exist20 June 2012 Venkatesh Umaashankar
  3. 3. Can we do better?  Recruitment is always driven by what is available in the job market.  Can we derive the multiple good job specs for a specific job, from the job market itself rather than limiting it to specs given by an expert?  Can a expert refine the job specs based on what is available in the job market?20 June 2012 Venkatesh Umaashankar
  4. 4. Leadership Positions in a fast growing analytics product firm – Job Specs  1. 6 – 10 years of experience  2. Analytics / predictive modeling experience  3. R, SAS, Weka, Java, Python, Matlab  4. Solve any 1 case study of your choice from www.kaggle.com and send us the way you solved it .20 June 2012 Venkatesh Umaashankar
  5. 5. Who you may end up hiring  B.E/B.Tech CS Major, 4 yrs, employers: IBM, SAS, Team lead, skillset: R, SAS, Java, C++, Weka, matlab. KDD cup participant  MBA, 10 Yrs experienced, Startup experience, Business Intelligence and Analytics head  M.Tech, IIT, 3+ years as software developer, MTech dissertation in machine learning, few publications in machine learning. Current employer: Oracle  M.Sc Mathematics, Anna university, 5 years experience in SAS, regression and predictive modeling.20 June 2012 Venkatesh Umaashankar
  6. 6. How to do better  A recruitment expert can define a generalized domain for a particular job.  Look in the job market ( by scanning the resumes of the persons who have applied for the job or the resumes available in the database of a job portal) and identify multiple good job specs for the job that are available in the job market which satisfy the defined domain.20 June 2012 Venkatesh Umaashankar
  7. 7. Advantages  Even before interviewing the applicants, the recruitment panel can understand what the market has really got to offer them.  Identify, some or many apt job specs which are available in the market and are scarce.  Ignore, widely available or less suitable job specs and interview only the persons who match good job specs.  Refine the intial job specs and come up with multiple job specs, that are actually available in the market and are not just hypothetical.  No more surprises or disappoints during the interview.  Interviews can be arranged only if market really has some suitable people.20 June 2012 Venkatesh Umaashankar
  8. 8. Graphical Model for Leadership Positions in a analytics firm TOP university Big names in Current and Ex employers Startup experience Maths Major Open source software exposure CS Major MBA Data mining Experience 3+ Experienced Experince in production systems Kaggle.com publications Blue nodes are boolean, can take values Y or N Prospective  20 June 2012 Green Node is unary, can take only one value Venkatesh Umaashankar Kdd cup Candidate
  9. 9. Probabilistic Queries on the Graphical model  The graphical model that we saw in the previous slide is a Bayesian network.  We can consider all the applicants as prospective candidates and calculate the conditional probability distribution for each node, using the resumes of the applicants.  We can make probabilistic queries like shown below to  What is the probability of a finding a person who has 3+ experience, MBA, open source exposure and data mining experience.  What is the probability of person having startup experience, given he has data mining experience.20 June 2012 Venkatesh Umaashankar

×