Real Estate Prediction Analytics

When buying property, how do you know it's the right move? Even if you're projected a net positive, how do you know it's the best alternative? Our ML tool uses historical data to help decision-making.

Comment

What it does

This tool is built in Python so that with given input data (in .csv format), it uses XG Boost, a ML tool, to use this data to project future returns if you were to buy a property in a specific town based on the historical patterns of pricing and valuation of the median house in the area. Our input and training data used 25 year historical data pulled from Zillow, filtered to Texas' cities and towns. Using this data, our program predicts the 5 year CAGR for each town, the price level, and the 3 year momentum of the prices.

How we built it

Once we knew what sort of data we wanted to pull and knew what calculations we needed for this prediction tool to be successful, we leveraged LLM tools to develop the back-end processes, debugged and refined, to have this current project.

Challenges we ran into

One of our earliest challenges was figuring out what data to use and where to get it. Obviously, the real estate market is influenced by more than just historical prices. Factors like education, corporate investment rates, demographics, income, taxes, and macroeconomic factors influence the specifics of the housing market. However, the biggest issue we ran into was that in order to have a tool that would be able to be as granular as the town level, we needed all the data to also go into similar levels of granularity, and the data available was not clean, nor did it do this. For simplicity and time's sake, we chose to train our model based solely on the historical Zillow data to predict the future pricing and CAGR.

Another issue we ran into was our lack of technical programming knowledge. While our idea was strong, we needed to use LLMs to help us find and implement the right tools to be able to train a ML model to analyze historical data and predict pricing, and in doing so, we struggled with debugging code that we weren't as familiar with.

Accomplishments that we're proud of

We're proud of having a functioning tool that can predict the 5 year CAGR, selling price, and the momentum to tell us what our investment will return and what risk factors to consider. We think that this is an excellent tool that real-estate investors and agents can make good use of, primarily because of its strong mathematical, logical, and entrepeneurial.

What we learned

We learned about the basics of XG boost and ML tools, and the framework for how this process works. Coming in with limited coding knowledge, we didn't expect to create a functioning tool, but we were able to learn a lot and were able to familiarize ourselves with a lot of how the tool developing process works.

What's next for Real Estate Prediction Analytics

We want to expand this tool to be able to take multiple input data points and files for the other categories and be able to eventually take in and normalize all data so that we can have a more accurate and detailed predictor.

Built With

Share this project:

Updates