Ames Iowa Data
We will be looking at 2 different CSV files of features from housing reports in a csv file. This data is collected from Ames, Iowa and was got from Kaggle.
- Training Data
- Testing Data
Which features affect sale price the most? And which features do not affect it at all?
All of the major points for the data and description can be found out here.
Points of Interest
- When cleaning the data, I first tried to change descriptions into rating numbers but that seemed to have a negative impact on my model so I quicly switched to dummy columns that definitely improved it.
- On the null values, I used different columns to find the mean for similar houses. I believed this to be better than setting it to 0 or getting rid of the column.
- There are some outliers in gorund livin area that I will not fix because I believe it will be dependent on the neighborhood which I will pull into account as well.
- Linear regression and LassoCV models had very similar residuals which made me look into the distribution of true y.
- Decided to use power transform with LassoCV and it turned out a lot better, still changes to be made.
- After looking at the coefficents and getting rid of some unncessary features; the model worked even better.
- It seems there are a lot of different features that will affect the sale price of a house in Ames, Iowa. However, lot frontage had no impact.
- Above Ground Living Area seemed to have the most impact.
- Overall Quality also had an impact.
- Year Built had an impact.
- Runner up Overall Condition.