Business Context

HomeValueNow, a residential real estate platform, wants to use supervised learning for two related tasks: estimating a home's sale price and predicting whether a listing will sell above asking price. The team wants you to explain when to use regression versus classification, then build baseline models for both tasks on the same dataset.

Dataset

You are given a historical dataset of 120,000 home listings from the last 24 months across 12 metro areas.

• Regression target: sale_price (continuous USD value)
• Classification target: sold_above_asking (1 if sale_price > asking_price, else 0)
• Class balance: 41% positive, 59% negative for classification
• Missing data: 12% missing in hoa_fee, 8% in lot_size, 6% in agent_experience_years

Success Criteria

A strong solution should:

• Clearly explain supervised learning and the difference between classification and regression
• Build one regression model and one classification model using the same feature set where appropriate
• Achieve MAE < $32,000 for sale price prediction and ROC-AUC > 0.80 for above-asking prediction
• Show how evaluation metrics differ by problem type

Constraints

• Predictions will run in batch once per day for up to 200,000 active listings
• The business team needs interpretable drivers, not only raw predictions
• The solution should be simple enough to maintain by a small data team

Deliverables

1. Define supervised learning and explain when to use classification versus regression
2. Build a regression pipeline for sale_price
3. Build a classification pipeline for sold_above_asking
4. Compare metrics, thresholding, and business tradeoffs across both tasks
5. Recommend which model should be deployed first and why