Project goals were to create Automated Valuation model and Bot for Australian Real Estate Market.
Project Details
Initially, the task was to create an AVM model which will predict property prices as close as possible. Target was the Australian estate market and I started from data collection by combing APIs like https://developer.domain.com.au/ and scraping directly from real estate websites. After gathering much-needed data I created an XGBoost model which has outperformed other algorithms. My target was to improve R2 and I have reached around 84% which was a pretty good result for my use case.
As a next stage client shared an algorithm which was basically how a real estate broker will evaluate the property under consideration to yield the highest profit. Details were clear and after some time prototype was ready and I have started testing. But as you can imagine it required much-needed data. However, the alpha version all features and currently it is just a matter of having enough data to cover all combinations.