/Predicting-fit-of-a-candidate-for-a-job-using-tree-based-models

We use multiple Tree boosting models and compare their performance to calculate the fit percentage of a candidate for the job they apply for. Also try to handle categorical methods using various methods.

Primary LanguageJupyter Notebook

Predicting-Fit-of-a-candidate-for-a-job-using-tree-based-models

Gender Neutrality and Inclusion

Problem

Companies are expected to be equal opportunity employers. Their recruitment and selection process is expected to provide equal opportunity irrespective of gender identity or expression, religion, color, sex, age, physical or mental disability, sexual orientation, or any other basis covered by local law. Removal of bias and providing equal opportunity to all applicants promotes access to the widest pool of talent. XyX corporation recruits employees for fixed Job Codes every year. Basis the applications received last year for each job code profile, a ‘fitment %’ percentage was determined based on selections made. XyX followed a fair and equitable approach by personally looking at all parameters and determining the right fit. This year the number of applicants has multiplied and they are looking at an ML model to predict the ‘fitment %’ for the applications received

Task

Build a model that calculates the ‘fitment %’ & detects the factor that influences relevancy and making sure that factor does not introduce inequality and/or bias in the’ fitment %’ by appropriate feature reengineering. Submit a presentation explaining how your model’s predictions will be used by business leaders to analyse and enable an equal-opportunity and bias-free recruitment process.

The Code is structured as follows:

  1. EDA
  2. Feature Engineering
  3. Feature Selection
  4. Update BiasInfluentialFactor field to handle bias
  5. Model Training
  6. Model Testing

Models Used:

  1. XG Boost
  2. Cat Boost
  3. Light GBM