Skip to content

Latest commit

 

History

History
35 lines (27 loc) · 2.29 KB

README.md

File metadata and controls

35 lines (27 loc) · 2.29 KB
Raw

IowaHousePrices

The following repository contains code for predicting house prices in Ames, Iowa from the Kaggle Competition: https://www.kaggle.com/c/house-prices-advanced-regression-techniques

python codes: boxcoxplot.py - boxcox transformation and visualization function for feature engineering outliers.py - function that detects outliers used in feature engineering

The following is a description of each folder in the repository:

Feature Engineering

There are two parts: the first python notebook is only from one hot encoding variables and the second feature engineering file is for label encoding of categorical variables.

Data

The Data folder contains both featured engineered data sets:

  1. train_x.csv, train_y.csv, and text_x.csv all correspond to the first feature engineering notebook: FeatureEngineeringRound1.ipynb
  2. train_x2.csv, train_y2.csv, and text_x2.csv all correspond to the second feature engineering notebook: FeatureEngineeringRound2.ipynb

The best scoring Kaggle predictions are in this folder as well called:

  1. Final_Prediction_1.csv - Prediction run from Elastic Net and SVR Average. Scored 0.1169 on Public Leaderboard
  2. Final_PredictionBestScore.csv - Prediction run from weighted average of Elastic Net, SVR, KRR, and GBR Scored 0.1154 on Public Leaderboard which was in the top 9%

Individual Model Predictions

In this folder each model prediction was ran in a separate ipython notebook listed by title:

  1. Lin_Reg_nb.ipynb - all linear regression model hyper parameter tuning and predictions: Ridge, Elastic Net, Lasso, and Kernel Ridge Regression
  2. RandomForest.ipynb - random forest model tuning and prediction
  3. SVR.ipynb - support vector regression model hyper parameter tuning and predictions
  4. Boosting Models.ipynb - gradient boosting tree model hyper parameter tuning and predictions

Ensembling Predictions

  1. Ensembling_v1.ipynb - ensembling Averaged and Stacking models for feature engineering notebook: FeatureEngineeringRound1.ipynb
  2. Ensembling_v2.ipynb - ensembling Averaged and Stacking models for feature engineering notebook: FeatureEngineeringRound2.ipynb
  3. Ensembling_Weighted_Averages.ipynb - "Kaggle Hacking" average weight ensembling method

Data Visualizations and testing

This folder is an extra folder containing extra EDA and Data visualizations