I have implemented an application to predict home pricing. Purchasing one’s first home is a major milestone, and getting a good price is important. Depending on the state of the real estate market, the price of a home may vary by hundreds of thousands of dollars. Even now, as home prices have hit record highs in the US, it’s possible that prices will continue to increase from here. So, should a prospective buyer bite the bullet now, or wait? This is where my home price forecasting application will come in. Based on historical pricing data, my application will leverage EvaDB to predict future home prices based on user-specified criteria. This will enable buyers to make a better-educated decision.
Note: This application currently only works to forecast housing prices in Connecticut. I planned to develop an application that would forecast prices for the entire USA, but I was unable to find a suitable dataset. In order to support other states, a larger dataset must be found.
First, you must stand up the backing PostgresSQL database. The included docker-compose.yml file does this for you. To use it, you must install Docker. Then, run
docker compose upfrom the project's root directory.
Then, run the application with the following command:
python3 forecast.py --town <TOWN> --type <TYPE> --frequency <FREQUENCY> --horizon <HORIZON> --start <START> --neural <T/F>Parameter descriptions:
town: town to forecast pricing for. valid values: any town in Connecticuttype: what type of property to forecast prices for. valid values: "Condo", "Single Family", "Two Family", "Three Family", "Vacant Land", "Commerical", "Apartments"frequency: prediction granularity. valid values: D, W, M, Y (day, week, month, year)horizon: how many predictions to make (e.g. 3 for 3 months when prediction frequency is set to M)start: the start date for the predictions (format: YYYY-MM-DD, default: today's date)neural: should we use NeuralForecast over StatsForecast? (T/F, default: F)
NeuralForecast will generally provide improved predictions over StatsForecast, at the cost of speed.
After running
python3 forecast.py --town Stamford --type "Single Family" --frequency W --horizon 3you should get the following result:
Generating forecast for Bridgeport:
Property Type: Single Family
Prediction Frequency: M
Prediction Horizon: 12
Start Date: 2023-11-25
Library: StatsForecast
| | Date | Price |
|---:|:-----------|--------:|
| 0 | 2023-11-30 | 238308 |
| 1 | 2023-12-31 | 238308 |
| 2 | 2024-01-31 | 238308 |
...