noellebr / homeprices

A project to predict sale prices for a Kaggle Competition

HomePrices

This project, created for MSDS 6372: Statistics II investigates obvious as well as non-obvious factors that influence sale prices of houses in Ames, Iowa. This project was submitted to the Kaggle Competition. By Noelle Brown and Thejas Prasad.

• Languages Used: R and SAS

Findings:

When ranking our models by both the test average square error as well as the Kaggle score, our custom model built off a backwards elimination technique provides us with the lowest ASE (test) and Kaggle score. The model included obvious features such as overall quality and condition, the year the home was built and remodeled, the total square footage, and the total number of baths. However, it also included less obvious factors, such as heating and air, the square footage of the unfinished basement, the screen porch area, and the wood deck area. Further research can be done on this project using more advanced regression techniques to come up with an even more predictive model.