Problem Statement

A company hires 950 individuals from various colleges across the country. After a year, they do a comprehensive performance appraisal for all these individuals. The appraisal is based on their on-job performance: a combination of objective metrics met by them and subjective rating by their managers. Each candidate is rated as a best performer (bp), mid performer (mp) or a low performer (lp).

During the hiring process, the company had taken these candidates through 15 different tests and an incredible set of profiling. Yet, they landed up with mid and low performers! They want to go through the various profiling parameters and hiring test scores to come with a predictor based on these parameters which will help them minimize mid and low performers. They will use this predictor in their next hiring process, so that they can hire candidates who will end up as high performers in their company.

One famous professor from a business school told the company officials that their problem can be solved by some funky computer scientists who call themselves Machine Learning Experts. They have engaged you as their Machine Learning Expert!

They have the following requirements:

  1. Multiple predictors

    Their team of recruiters have a right wing and a left wing; some of them call themselves the conservatives and some the liberals for unknown reasons. They have different philosophies of hiring:

    The Conservatives: They are happy to lose a few best and mid performers but cannot tolerate low performers at all.
    The Liberals: They are happy to let a few low and mid performers come into the system, but do not want to lose any best performer.

    For now, we do not know who won and who lost. The machine learning expert offered to give each of them a different, their very own predictor.

  2. Understanding the predictor

    The company is not interested in a black box predictor only, but wants to understand what makes someone a best performer at their company and what does not. For instance, one of the HR managers has a strong belief in astrology (Refer Disclaimer) and thinks candidates with birth in certain months are high performers! Can your predictor tell whether this is true or not. Or otherwise, can your model tell what variables really make someone a high performer at their company?

    They want the machine learning expert to provide them some insights rather than just handing them a programming function they can run on each candidate. Do you think you can help this company out?

Data Sets

There are two separate sets of data. In the first data set is the train data set. All the input parameters for each candidate have been provided together with the output as best performer, mid performer and low performer. The second smaller set is the validation set, wherein only the input parameters have been provided.

Once you register, the problem data sets shall be emailed to you within 72hrs.


The current problem or predictors do not influence any company's hiring policy. They are here completely for an academic purpose and with a goal of developing better computer engineering and science. Any mention of particular groups, mindsets in a company or belief in certain hiring principles is simply from a creative standpoint. It is not any particular company's policy or a policy advised by Aspiring Minds.