Melbourne, 2016-2018 sales

What would it sell for?

An XGBoost model trained on real Melbourne sales, tested only on sales it never saw. It gives a range, not just a number, and tells you how it compares to the plain suburb median.

Type
More detail (optional)
Left blank, these fall back to the typical value from the training data - the same thing the model saw when a listing left them out.

Pick a suburb and press Estimate.

How good is it?

Trained on sales up to October 2017, then tested on - later sales the model had never seen.

Median errorWithin 10%Average $ error
Suburb + type median
This model

The 80% range held the real price for of those future sales. Everything runs in your browser: the trees are exported to JSON and scored in plain JavaScript, checked against Python in the test suite.

Code and write-up · rishabhray.me · Data: Melbourne Housing Market by Tony Pino, CC BY-NC-SA 4.0