Real Estate Price Prediction

Berkeley SAAS CX Fall 2020 Kaggle Competition

Ming Fong and Yifan Zhang

Winning solution

Predicting real estate sale prices using property data.

The code used for our final submission can be found in final_submission.ipynb.

Data

Data can be downloaded from the Kaggle competition data page. In the repo, data is in the /data directory.

There are 3 data files:

• data/test_features.csv
• data/train_features.csv
• data/train_targets.csv

output/sample_submission.csv is an example of a file that is ready to submit to Kaggle. There are two columes: id and SALE PRICE.

Kaggle Link

https://www.kaggle.com/c/saas-2020-fall-cx-kaggle-compeition

Notes

Building codes: https://www1.nyc.gov/assets/finance/jump/hlpbldgcode.html

TODO

• remove outliers
• check negative price predicitons -Check if building or tax class changes
  • could mean redeveloped housing
  • add column "classChanged" - 1 if yes, 0 if no
• Check if apartment number is present
  • add column "hasApartmentNumber" - 1 or 0