Queens Price Prediction: Random Forest Cut RMSE by $6.5K

PUBLIC-DATA / ACADEMIC ANALYSIS

Business question How accurately can Queens apartment sale prices be predicted from available transaction and property features, and which model provides the strongest out-of-sample performance?

Sample & method Analyzed 521 Queens apartment sales and compared ordinary least squares, regression tree, and random forest models in R. The evaluation focused on predictive accuracy rather than training fit.

Evidence • Random forest achieved R² = 0.79. • Its RMSE was $6,500 lower than the OLS baseline. • The tree-based model’s advantage suggests that nonlinear relationships and feature interactions matter for this pricing problem.

Uncertainty The sample is limited to 521 Queens transactions and may not generalize to other boroughs, property types, or market periods. R² and RMSE measure prediction, not causation. Before deployment, I would add time-based validation, leakage checks, and model-drift monitoring.

Business decision USE THE RANDOM FOREST AS DECISION SUPPORT, NOT AN AUTOMATED VALUATION. It can prioritize manual pricing reviews and flag potentially mispriced listings, while human judgment handles unusual properties and changing market conditions.