As a group project for the CDIPS 2017 Data Science workshop, we used machine learning techniques to analyize Kaggle's House Prices data. We informed feature engineering by inspecting data for distribution types and required transformations, along with looking for some basic underlying relationships between features and house prices. To determine the best predictors of house prices we compared linear models, elastic nets, LASSO regression, Random Forests, and XGboost to choose feature sets. We also compared limited feature sets before and after modeling to determine whether adequate predictions could be made using more limited information. Our included short presentation shows comparisons of models and feature sets.
Information about the CDIPS workshop and Python ML tutorials provided: http://cdips.physics.berkeley.edu/2017-cdips-data-science-workshop/ http://cdips.physics.berkeley.edu/tutorials-2017/
The Kaggle house prices challenge also contains several tutorial kernels of interest for learning ML / Data Science in Python: https://www.kaggle.com/c/house-prices-advanced-regression-techniques#tutorials