A forecasting research track that combined transaction history, macro signals, and richer geospatial context to model Singapore property-price movement more rigorously.
Outcome: Developed forecasting models that moved beyond simple headline variables by incorporating broader market, amenity, and geospatial context.
Housing-market forecasting is often reduced to a few popular variables, even though price movement is shaped by a wider system of supply, financing, neighbourhood context, and transaction behaviour.
I developed a modelling track that combined housing transactions with macro signals, supply context, and geospatial features to test how much more explanatory power could be gained from a richer input layer.
The work drew on multiple modelling approaches rather than relying on a single method:
An important part of the research was feature engineering. Instead of treating location as a single proximity variable, the work incorporated broader amenity and neighbourhood context so that the models reflected more of how urban housing markets actually behave.
The value of the project was not simply in producing a forecast number.
It helped build a more disciplined way to think about prediction in Singapore real estate: use models to test structure, compare signal strength, and understand uncertainty rather than overstate confidence. That framing still shapes how I approach market scenarios today.