Predicting typhoons and visualising their impact for Red Cross 1


As technology grows and becomes more and more a part of everyday life, we are starting to truly understand how to utilize it for the good of society.  Using data for social change is a relatively new trend.  However, it is picking up steam and many organizations are looking into new innovative ways to use this valuable tool for the greater good of humanity.

Many developmental organizations like the Red Cross are harnessing their data to implement new and improved methodologies in combating natural disasters and other social problems. However, these developmental organizations tend to lag behind the for-profit world in capturing data and using analytical tools to make better decisions. “Most nonprofits don’t have the resources to bring this expertise in-house,” says Lucy Bernholz, visiting scholar at the Stanford Center on Philanthropy and Civil Society. “But nonprofits have tons of data. They are a critical part of our collective data ecosystem.”

Red Cross and DataMission teamed up for Hackathon on the 20th of May. The Red Cross has a team, whose mission is to shape the future of humanitarian aid by converting data into understanding. The event focused on two main challenges about typhoons in The Philippines. The first challenge was to create a machine learning algorithm to predict the damage a certain typhoon has made, right after it has struck. The second challenge was on the visualization and storytelling of the model and its predictions. This to make it accessible to managers, relief teams and/or civilians.

In total there were seven teams. Five of the teams focused on the machine learning challenge and the other two teams worked on the second challenge.  Team negative R was the winner for challenge one. Their approach was similar to those of the other teams but their predictions were better. They concluded that the data for the number of damaged households was sometimes higher than the number of actual households. This surprising fact comes because the number of households was measured in 2010, but the number of damaged households in 2015. They started with simple linear models, to find out which parameters were important. They later filtered out the correlated inputs. Their result was R squared of 0.68. More info about the other approaches tried by the other participants is here.

Marco of Red Cross then said that none of the models developed in the hackathon was as good as the Red Cross model. Knowing this is very valuable to the Red Cross, because they now know that many other approaches, which they didn’t try, don’t improve their model. Marco emphasized that the hackathon teams only spent one day on this, while the Red Cross model was the result of a team effort over several months.

Then it was time for the visualization teams. For the data science teams there was a clear rule to find the winner. But the two visualization teams were scored by the public, based on how well they did on 1) visualization 2) storytelling and 3) how they communicate accuracy. The stakeholders have to be convinced of the accuracy of the model before they base their actions on what the model says.

Team two in the visualization group was the winner of the group. They used two axes to define the users of the visualisation. For instance, do they have high data literacy, or low? Do they require information from the entire country or just a few provinces?

They use the map to tell a story:

  • What happened? (i.e. what was the wind speed)
  • What is the damage? (i.e. how many houses were destroyed)
  • Who was affected? (this using a methodology based on the inform index)
  • Which agencies are on the ground and can help?

To give a good overview of the data the showed it not only on a map, but also as bar charts that showed the number of affected houses per municipality. After all, some municipalities that have the same size on the map can have a completely different number of houses. They also show the timeline at the bottom. They thus show not just what is happening on the ground, but also when. Just like Ushahidi they want to involve citizens in measuring the damage, to get timely results.

It was a good and fruitful hackathon. Many thanks to all who sacrificed a day of their weekend to help! Also thanks to our sponsors (GoDataDriven & Big Data Republic) who both organise interesting meetups.

Finally, the typhoon season in The Philippines will unfortunately start again in August. really would like to have then an improved model. Don’t hesitate to contact them if you want to help them with this, or with one of their other projects!

Leave a comment

One thought on “Predicting typhoons and visualising their impact for Red Cross