Case Shiller Prediction

I’ve worked in the real estate industry for most of my life. My work has been pragmatic though –construction, property management, labor. I only recently gained an interest in how the broader real estate market moves, and further how data can be leveraged to gain insights about the market.

Project

As a final project for CS 229: Machine Learning, Matt Wolff, Laywood Fayne and I compiled economic data and used it to predict the Case Shiller housing index in 20 major US cities. Find the full source code for the project here.

Method + Results

I implemented the Vector Autoregression algorithem on our data. VAR is a statistical model that captures the relationship between multiple variables, and uses these values to model future trends. More about the VAR model can be found in our final write-up.

We ended with some nice graphs predicting future values of many features. See below for our predictions vs actual values in San Francisco: