FairResale

Inspiration

My inspiration is from the million-dollar HDB resale headlines that grabs attention, but these articles don’t necessary tell buyers which towns and flat types fit their own budget. These million-dollar transactions made known can possibly scare and deter potential buyers. So, intention is to turn a market-wide concern into a practical starting point for a home search.

What it does

FairResale helps prospective buyers compare HDB towns against a chosen budget and flat type. It will show recent prices, the share of comparable past sales below budget, transaction counts, and a six-month forecast range.

How I plan to build it

Databricks will ingest HDB resale transactions and SingStat income data, clean and summarize them into Delta tables, and then track a town-level forecast against a simple baseline with MLflow. An integrated Databricks App will let buyers enter a budget and explore the results.

Challenges expected

Town and flat-type sales can be relatively sparse, which may make some comparisons and forecasts unreliable. Sample counts and uncertainty, and label low-volume results will be shown.

What I aim to accomplish

Aim is to give buyers a clearer shortlist of towns to investigate. The pilot target is for 8 out of 10 test users to identify three plausible towns in half the time of a manual search.

What I hope to learn

To learn whether budget-based comparisons help buyers narrow and aid their search, and whether a town-level forecast can add useful context without overstating what past transactions can predict.

What’s next for FairResale

If selected, I'll build and test one end-to-end Databricks workflow: ingest the data, validate the summaries and forecast, then rehearse the buyer search with mock users.

Built With

• data.gov.sg
• mlflow
• sql