Hunterb9101 / Kaggle-Playground

Kaggle Playground

Problem Statement

Model development time for a Kaggle Playground competition is extremely limited. Thus it is in our best interests to follow a design pattern when trying to tackle these problems.

Key Model Stages

• EDA

  • Check for nulls
  • Outlier Counts

• Data Processing

  • Feature Engineering
    • Standardization (For Linear Regression)
    • Feature Addition
    • Outlier Removal
    • Fill NAs
    • Categorical Encoding
    • PCA (Hyperparameter dependent)
    • Binning
  • Feature Selection
    • VIF (Linear Regression)
    • Correlation / Variation Pruning
    • Recursive Feature Elimination (Hyperparameter dependent)

• Model

  • Model Hyperparameter Optimization
  • Model training
  • Model Analytics
    • Metric scores
    • Top features

• Postprocessing

  • Round Final Inputs
  • Model Stack analytics
    • Metric Scores
    • Model Correlations
  • Model Stacking

Key Notes

- EDA: No output