(rank as of this commit)
Files:
- HousePricingDataProcessing: Pre-processing of features.
- Used this notebook as a coarse guide and learn what steps can be taken.
- Saves training and test(for submission) data after feature engineering
- SearchFeatureSelector: Runs a search over bayesian regression using different numbers of important features for fitting
- HousePricingV3: Implements the House pricing regression
- reads processed data
- handles categorical data - one hot encoding
- applies multiple regression techniques
- linear regression (with and without log on target)
- Random Forest
- Lasso
- NN
- Bayesian Regression
- XG Boost
- Try stacking of these models
- Use result from SearchFeatureSelector to get the best relevant features and use them for fitting bayesian (since it was giving the best results)
- run on all data before generating final submission
- You can download the file with actual correct values on test data OR remove lines of code with "full-score.csv" in all files.