Futurenest

Inspiration

With the Palisades fires still ongoing here in SoCal, we've already seen major devastation to various houses and other properties. This of course has caused major outcry and fear over how increasing climate-caused disasters will have an affect on the most owned investment asset by Americans.

What it does

FutureNest is a property investment prediction tool designed to help users glance at how Wildfire risk will continue to affect the housing market in various counties across the United States. The application provides a sleek, user-friendly interface that allows the user to enter their desired county and state, which the application then uses to predict future pricing for the next ten years. While definitely not being the first housing price forecasting tool, FutureNest's custom-trained ML model using wildfire risk data carves a specific niche for determining long-term investment risk for both businesses and families alike.

How we built it

FutureNest uses data collected from Zillow's neural-network trained ZHVI housing data along with data from FEMA's wildfire risk index to train a hand-picked random forest regression model on over 8500+ feature sets of various counties across the US. The Application itself is hosted through a flask server deployed to Vercel, making use of HTML/CSS/JS for a sleek & simplistic front-end experience (along with the integration of chart.js for plotting)

Challenges we ran into

Collecting and wrangling data for a model as complex as this in the span of 24 hrs was... difficult to say the least (and very limiting!)
Integrating both front & back-end communication, especially with finding a way to incorporate the model into the back-end architecture

Accomplishments that we're proud of

Produced a pretty accurate (r2 of ~0.65) random forest regression model on a cobbled-together dataset collected within the span of 24 hrs -A lot of us were new to the technologies we were using. Whether it was Flask, Vercel, APIs, or scikit-learn, we were still able to use these technologies to incorporate them into our program.

What's next for Futurenest

Focus on retraining the model on a more feature-packed dataset (currently limited by only using ZHVI and FEMA's risk index)
incorporate risk scoring into the application: giving the average user a more general idea of how risky property in a specific county will become