HDBBQ

Team Members of Team 'HDBBQ'

Thamudaya Win Berry, NUS, Business Analytics, Year 4, [email protected]
Leow Ying Jie Bernie, NUS, Business Analytics, Year 4, [email protected]
Nagam Varshini, NUS, Business Analytics, Year 4, [email protected]

Inspiration

Over the past few weeks, stepping outside in Singapore has felt a little different. The weather has been hot, the skies have been hazy, and suddenly we found ourselves checking the PSI before heading out. But that got us thinking: we all experience the same weather, but do we all face the same risk?

Probably not.

A young adult living near an air-conditioned mall might cope very differently from an elderly resident living in an area with fewer nearby places to cool down. And yet, most weather apps tell us the same thing:

It is 34°C. The PSI is high. Drink more water.

Useful? Yes. But we wanted to know what comes next. Where are the people who are actually most at risk? And if we could identify them, what could we do about it?

What it does

HDBBQ looks beyond temperature alone. We bring together live environmental conditions with information such as elderly population density and access to cooling spaces to estimate a Heat Risk Score for different areas in Singapore.

Think of it this way: Hot weather + vulnerable residents + poor cooling access = higher risk. On our dashboard, users can see Singapore's current heat conditions on a map and quickly spot areas classified as Low, Medium or High risk.

Click into an area and HDBBQ explains why its risk is high and suggests what could be done from monitoring the situation to prioritising elderly outreach or improving access to cooling spaces.

And instead of stopping at “this area is high risk”, we wanted to ask: What if we actually did something about it?

Our simulation lets users explore how adding cooling infrastructure to a high-risk area could potentially change its risk profile. So HDBBQ isn't just about finding hot places. It's about finding where heat could hurt the most and what we can do next.

How we built it

We envision HDBBQ as an interactive dashboard built using Databricks. It would bring together environmental data with demographic and infrastructure datasets to help users understand heat risk across Singapore.

The proposed pipeline is: Data ingestion → Cleaning and transformation → Risk calculation → Area classification → Dashboard visualisation

We ingest environmental readings such as temperature and rainfall, then combine them with information on elderly populations and nearby cooling infrastructure.

From these features, our model produces a Heat Risk Score for each area. We then translate the score into something much easier to act on:

Low which would mean continue monitoring
Medium which would mean prepare and watch vulnerable groups
High which would mean prioritise intervention

Finally, everything comes together in an interactive dashboard where users can explore the map, understand what is driving the risk, and test possible interventions.

Challenges we ran into

One of our first challenges came before we even started building: figuring out how all the data would fit together.

The datasets looked straightforward individually, but once we started researching them, we realised they operate at very different levels. Weather readings come from individual stations and can update every few minutes. Elderly population data is tied to residential and geographic areas. Cooling facilities are individual locations on a map.

We also found that some datasets used different planning boundaries and reference years. For example, our initial data used 2025 planning-area boundaries, while some of the demographic data was based on the 2019 boundaries. We therefore had to decide which boundary system to standardise around so that the datasets could be joined consistently.

So we had to think carefully about questions like:

How do we map a weather station reading to a planning area?
How do we combine live data with datasets that barely change?

Working through these questions early helped us realise that designing the data pipeline was just as important as choosing the model at the end of it.

Accomplishments that we're proud of

Before submitting our idea, we didn't want our architecture to exist only as boxes and arrows on a slide. So we tried out the Databricks features ourselves to make sure the key parts of our proposed pipeline were actually feasible.

Not everyone on our team came into this equally familiar with Databricks, which actually made this part of the process especially valuable. We got to explore the platform hands-on, understand how the different pieces fit together and identify where our original assumptions needed to change.

It also gave us much more confidence in the solution we were proposing. Instead of simply assuming that the pipeline would work, we were able to test it ourselves, understand how the different pieces connected, and feel ready to build it. For us, that was a small but meaningful win before the hackathon even began.

What we learned

One thing we learned was that it's easy to get excited about what the technology can do before asking what the user actually needs it to do. At first, there were plenty of features we could imagine adding such as adding more datasets, more models, more predictions. But as we worked through the idea, we kept coming back to one question: “So what?” If an area has a Heat Risk Score of 82, what does that actually mean to the person looking at the dashboard?

That changed how we thought about HDBBQ. Instead of trying to show as much data as possible, we started thinking about how each piece of data could lead to a useful decision. A temperature reading becomes a risk signal. A risk score becomes an explanation. And an explanation should eventually lead to an action. That was probably our biggest takeaway. Good technology isn't just about producing an answer, rather it's about making that answer useful to someone.

What's next for HDBBQ

Our next step is to turn the pipeline we've planned and tested into the actual HDBBQ product. We'll start by connecting the environmental and demographic datasets, building the data pipeline in Databricks, and developing our first Heat Risk Score. From there, we'll bring the results to life through the interactive Singapore map and build out the Low / Medium / High risk classifications and recommendations. Once the core experience works, we'll move on to the features we're most excited about (the cooling infrastructure simulator and eventually, next-day heat risk forecasting).