Before the estimates are calculated, the housing index are generated using either hedonic or repeated sales models.
The underlying methodology behind RealValue consists of two methods, comparable and machine learning (XGBoost). Both methods use past transactions albeit used differently. Past transactions are first readjusted to the current times using the housing index generated.
Comparable method references past transactions of similar properties. A pre-defined hierarchy of criteria are used to determine similar past transactions. For condo and HDB, recency, location and floor area are used. More criteria are used for landed. Adjustments are made based on the differences in the properties.
For transactions that do not have comparable transactions, we rely on machine learning method. After extensive research, XGboost is the chosen approach due to model performance and state of the art practices. XGboost acts as a highly advanced decision tree, scrutinizing data to discern intricate patterns, similar to recognizing everyday trends. The model is trained on all transactions in the past 5 years to predict current transaction price.
The final valuation will be a combination of comparable and machine learning model, if comparable is available, else it will be purely machine learning model.
For more information about RealValue, visit https://tech-rea.com/realvalue
If you require assistance, please contact our customer support at +65 9751 2217 or help@realestateanalytics.sg.