Comparing property-price models with broader and narrower sets of neighbourhood amenities in Singapore.
In brief: Compared amenity feature sets for property-price modelling; this summary does not report validation metrics.
Does a broader set of neighbourhood amenities improve property-price modelling beyond a few proximity measures, such as distance to an MRT station or a school?
I combined property transactions from 1990 to 2023 with amenity data across dozens of categories, including transport, schools, malls, and other points of interest.
SQL and Python were used to prepare and join the data. The analysis tested correlation-based approaches, LASSO, Random Forest, and LSTM variants to compare different feature sets and modelling methods.
The project compared broader and narrower amenity feature sets. This summary does not provide error metrics or a validation split, so it does not quantify predictive improvement or establish out-of-sample superiority. The project informed my later work on geospatial feature engineering and neighbourhood context in housing analysis.