Scraping hemnet and building models. Hemnet is a property listing website. These scripts can scrape both current listings and recently sold properties, and then can model the sale price.
The data is for apartments in northern stockholm.
I used catboost as the regressor, it turned out OK, this shapley plot is probably the most interesting:
Who would have know that floor space would have been a good predictor. Perhaps the most interesting things to be read from this is the jump in price from 2 to 3 bedrooms. The model could be improved by adjusting for the floor space and seeing if there are any standout factors.
As it is now, you probably could use this model to try and find good deals on hemnet, but probably not that successfully.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent