https://www.kaggle.com/c/zillow-prize-1
In this million-dollar competition, participants will develop an algorithm that makes predictions about the future sale prices of homes. The contest is structured into two rounds, the qualifying round which opens May 24, 2017 and the private round for the 100 top qualifying teams that opens on Feb 1st, 2018. In the qualifying round, you’ll be building a model to improve the Zestimate residual error. In the final round, you’ll build a home valuation algorithm from the ground up, using external data sources to help engineer new features that give your model an edge over the competition.
The code is a mutli-level model using several optimized base models that are then stacked with a meta model. The final goal is have 2 or three meta models and combine then with either arithmetic and/or geometric averaging.
Base models:
- Random forest
- ADA boost
- Gradient boost
- KNeighbours (for 1,5,10)
Meta model:
- Linear regression
Requirements:
- Python 3.5+
- Sklearn 0.2
- Numpy 0.2
- Panadas
To run just type "python run.py". There is an optimisation flag when the model is called that can be set to run the find hyper paramaters function but be warned it will take some time.