I created a pipeline and model for a housing prices with a dataset of Cook County, Illinois.
The dataset consists of 500,000+ observations and 62 features.

The Data Pipleline

For each feature below, I cleaned the data by removing invalid data points as well as not including any outliers in the analysis and performing any transformation operations if needed. I first worked on the response variable, Sales Price. Click on image below to see the cleaned Sales Price distribution with a log transformation.

To include include the roof material into our predictor variables, I one hot encoded each of the different materials below, turning each into seperate indicator variables I added to each column. Click below on each box plot to see the distribution of their respective indicator variables.

From the description of each housing listing, I also used regex to extract the number of bedrooms in each house, below is a violin plot showing the relationship between number of bedrooms and sale price

Also, I grouped houses that shared neighborhoods and found the top 3 most expensive neighborhoods using pandas, and created an indicator variable for it called "in_expensive_neighborhood".

I then also took a look at the correlation between the area of each house and the sale price through a joint plot.

The Model

After running the data through the pipline functions above, I then split the data into training and testing data, 204792 used for the training set and 68264 used for the test set.
I constructed a linear classifier with scikit learn using existing features and the newly generated ones, with a Residual plot below.