🏡 Home Sales Prediction Challenge

Can you predict the number of home sales for the next 12 months?
This challenge puts your data skills to work using real estate data, a time-tested model, and your favorite LLM (ChatGPT, Claude, Gemini, etc.) to improve results.

You’ll get:

  • 📧 Starter Python scripts via email

  • 🧠 A working time series model

  • 🎯 A performance benchmark to beat

🎯 Your Goal

Use your own dataset to forecast monthly home sales for the next 12 months.

Start with a Holt-Winters model and improve on:

✅ Instructor Benchmark:
R²: 0.55
MAPE: 27.19%
Forecast: 79 home sales over 12 months
Estimated Range: 58 – 100 homes

🗂 What You’ll Do

Step 1: Convert Your Data (standalone.py)

This script reads your raw Flexmls data and creates a monthly summary of closed sales.
You’ll use the output (monthly_sales_summary.csv) as input for the forecasting model.

📌 Make sure to update the script to match your file name and column structure.

Step 2: Forecast Home Sales (flex.py)

This script loads your monthly summary file and uses a Holt-Winters model to:

  • 🔮 Forecast the next 12 months of sales

  • 📈 Report R² (accuracy score)

  • 📉 Show MAPE (average error)

  • 🔁 Print an estimated range of home sales

It’s a complete time series forecasting model — ready to run and improve.

🤖 Improve with an LLM

Once you’ve run the starter model, ask your LLM something like:

“My instructor gave me this Holt-Winters model in Python to forecast home sales.
It works, but I want to improve the R² score to 60% or higher.
What changes can I make? Should I try SARIMA, Prophet, or add macroeconomic indicators like mortgage rates or inflation?”

Your LLM will guide you through alternate models, new features, or tuning parameters.
Copy and paste its suggestions into your script and keep testing!

📧 You’ll Receive:

You’ll get the following Python files by email:

  • standalone.py – for summarizing monthly sales

  • flex.py – the Holt-Winters forecasting model

If you don’t receive the files or need help getting started, contact me.

💡 The Challenge

  1. ✅ Use standalone.py to generate your monthly summary

  2. 🔧 Run flex.py with your own data

  3. 🧠 Use your LLM to improve it — tweak the model, add features, or try a better algorithm

  4. 🥅 Try to beat my benchmark — aim for 60%+ R²

Good luck — and happy forecasting!
(And remember: debugging is part of the job 😄)

Spread the love