Loading, please wait.
Model transparency
An inspectable regression baseline, with its validation results and limitations in the open.
Training properties
50
Supplied labelled reference sample
Cross-validated MAE
7,107
Mean absolute error · source price units
Cross-validated RMSE
10,214
Larger errors receive more weight
Cross-validated R²
0.9835
Coefficient of determination, not accuracy %
| Feature | Coefficient | Observed range |
|---|---|---|
| Floor area | -102.92 | 980 – 2,400 |
| Bedrooms | -8,447.78 | 2 – 4 |
| Bathrooms | 7,436.93 | 1 – 3 |
| Year built | 112.38 | 1978 – 2012 |
| Lot size | 39.3 | 4,400 – 10,500 |
| Distance to centre | 16,810.21 | 2.1 – 8.2 |
| School rating | 21,062.08 | 6.5 – 9.1 |
| Intercept | -299,786.61 |
Coefficients describe fitted partial associations. Strong correlation between features can make individual signs unstable. They do not establish causal effects.
Pooled out-of-fold predictions from five-fold shuffled cross-validation. Each property is evaluated by a model trained without that property. The serving model is subsequently refitted on all labelled rows.
Mean-price baseline MAE
70,580
Compared using the same held-out folds
Model version
linear-regression-474a339a35c1
Trained (HKT)
26 Oct, 12:04 am
Dataset SHA-256
0e36c6224e1f6fb97c308c9dbe1d6da22d78d78055181067f8ac4c7155a4a726