coffeeboost / sml-grade-predictor

supervised machine learning Kaggle submission

Supervised ML Kaggle competion

(link to competition)[https://www.kaggle.com/c/cais-exec-team-in-house/overview]

Program Flow

• Data exploration
  • pairwise correlation matrix with RMSE
  • histogram for data distribution
• Data cleaning
  • remove grade == 0 data, probably noise
  • one hot encoding for categorical data
• Training
  • random forest classifier
  • XG Boost
• Testing
  • 70% train, 30% test
  • cross validation
• Tuning
  • grid search hyperparameter tuning
  • bruteforce search
  • feature importance plot
• Improvements
  • reduce columns to save training time

Result

top 3