Week 8

02/02/26

Another feature that we added was scraping images from listings to display on the final quote page. This was simply implemented by parsing the html and searching for real images. This feature helped give a more professional look to our final quote page.

We also built a simple machine learning model using linear regression to integrate into the website and act as a placeholder while we worked on the finished model. This allowed us to essentially finish the main flow of our website and allow users to get a predicted premium with our placeholder model.

Machine Learning

From the original 59 features we thought of, we sought to find which values are significant and which arent as part of building our final ML model

To decide this we used P values and coefficients.

These were all the features we found to be insignificant and as a result they were dropped.

We also used Linear regression to help us figure out which model will be the best to train our data with, here are the results.

We found that XGboost was clearly the best choice for our data.