Property Pulse

Logo
Home Page
Search by ROI
Search by Income
City Details

Inspiration

Real estate values are influenced by various economic factors, including cost of living, job availability, and domestic migration. Traditionally, the impact of these variables is assessed over extended periods. Our project seeks to address this lag in housing price predictions by leveraging near-real-time insights derived from moving patterns. U-Haul rental prices provide valuable data that reflects the dynamics of population movement. For example, a city experiencing an influx of residents will generally show lower rental costs for trucks departing the area compared to those arriving. This discrepancy can signal heightened demand for housing in the target city, forming the foundation of our analysis.

What it does

Property Pulse regularly updates a precise projection of the average real estate value in United States cities by implementing a neural network trained on web scraped U-Haul rental prices and other frequently used predictors.

How we built it

The project began with the development of a web scraper to extract U-Haul rental prices for specific routes. We supplemented this data with additional datasets, including housing prices, income levels, and unemployment rates, sourced from the U.S. Census Bureau API. By integrating diverse features and over 20 years of historical data, we trained a neural network capable of forecasting average real estate values effectively. To facilitate user engagement, we created a Django-based web application that enables users to identify cities with the highest predicted returns on investment in real estate. The application includes various selection interfaces to streamline the decision-making process.

Challenges we ran into

One of the biggest challenges was collecting data using a web scraper because there were so many unexpected issues that occurred. In addition, accumulating a significant number of datasets made it hard to merge all of them together while still keeping the integrity of the data intact. Integrating the neural network model statistics within the Django framework posed problems with indexing some of the data.

Accomplishments that we're proud of

We were proud of finding a machine learning model that could accurately employ our dataset. Additionally, it was exciting to successfully add a typewriter effect to our homepage. We are also glad that we were able to finish a project within the timeframe.

What we learned

We learned about the libraries (BeautifulSoup and Selenium) and technical knowledge (XPATH, HTTP requests/responses, etc) required to web scrape a site. We also got experience with creating a web application using Django.

What's next for Property Pulse

We hope to expand our project by including more cities into our model. By working with a greater amount of data, the results will also be more accurate and responsive to changes in prices. We would also like to make our web application more user-friendly and add various features. Also, we weren’t able to web scrape as much U-Haul data as we would have liked to. By continuing to scrape rental price data, we can train the model with more data specific to rental prices so that the model can update its property value projections with greater frequencies and finer margins than a typical real estate estimator.