Link to leaderboard: https://datahack.analyticsvidhya.com/contest/machine-learning-starter-program-hackathon/#LeaderBoard

The aim of this hackathon was to predict if a trainee will pass the examination for a particular program.

Given dataset had demographic details about the trainee as well the course.

• Final model was a weighted average blend of bagged CatBoost and Xgboost models
• Both models were trained on seperate datasets
• For CatBoost:
  • User SelectFromModel from sklearn to obtain import features
  • Created interactions amongst these features
  • Fed categorical features as it is to model
• For Xgboost:
  • Created interactions among intuitively selected features
  • Frequency encoded categorical feats

• Intuitive label encoding of categorical features
• Droping "Age" feature which has considerable missing values