Property Details
Select property class or enter a postal code for valuation.
🔮 Projecting future valuation post historical transaction cutoff (Sep 2026)
Client-Side AI Engine • Public HDB & Private Non-Landed (Sep 2026 Data)
Select property class or enter a postal code for valuation.
Click on any marker across Singapore to inspect actual registered historical transactions and caveats.
This application clearly separates AI model predictions (XGBoost) from actual recorded transactions. Below are the primary sources and dataset cutoff dates:
| View / Tab | Data Nature | Housing Class | Source Agency | Records | Historical Coverage | Latest Trained Cutoff |
|---|---|---|---|---|---|---|
| Tab 1: Price Predictor | AI Prediction | HDB Public Flats | Housing & Development Board | 239,977 | Jan 2017 – Sep 2026 | September 2026 |
| Tab 1: Price Predictor | AI Prediction | Private Condos / ECs | Urban Redevelopment Authority | 71,538 | Jun 2021 – Jun 2026 | June 2026 |
| Tab 2: Transaction Maps | Actual Data | HDB Block Popups | Data.gov.sg (HDB) | 9,818 Blocks | Jan 2017 – Sep 2026 | September 2026 |
| Tab 2: Transaction Maps | Actual Data | Condo Popups (URA Caveats) | URA Data Service / Ryeo | 83,848 Caveats | Jun 2021 – Jun 2026 | June 2026 |
To ensure reliable, transparent valuations, this application maintains two distinct, independently trained XGBoost models rather than blending them together. Each is trained exclusively on its official agency dataset and validated on future sales it never saw — a chronological holdout, not a random split.
The app produces two kinds of estimate. A comparable valuation prices a property at the current market level, and the accuracy figures below measure exactly that. An index-adjusted forward valuation takes that result and scales it by an official published price index to reach a future date — the XGBoost model itself never forecasts, since decision trees cannot extrapolate beyond the years they were trained on. Projected estimates are labelled with the growth rate and horizon they assume.
Share of HDB price variance explained across all 26 towns.
Measured on a held-out future year.
Spatial, flat & transport features; max tree depth 8.
Instant lookup of town, lease commence date & coordinates.
Applied to project prices beyond the training data.
Share of private resale variance explained across districts.
Measured on a held-out future year.
Engineered features (project, tenure, lease, storey, segment).
2,431 non-landed developments mapped: condominiums, apartments and ECs. Landed property is not covered.
Applied to project prices beyond the training data.
This application operates on official Singapore public housing records and real estate caveat registrations:
Historical resale transactions from January 2017 onwards, published by the Housing & Development Board (HDB) via Data.gov.sg under the Singapore Open Data Licence.
Private residential caveats sourced from the Urban Redevelopment Authority (URA) Real Estate Data Service, via the Singapore Property Transactions Dataset by Ryeo (CC BY 4.0).
Explore the other open-source projects in this Singapore housing intelligence suite: