#HackHunger with UN World Food Programme and DataMission
For the past couple of months, the DataMission team has worked hard to organise its first hackathon on Saturday May 14 in Amsterdam. And we had a blast! A big thanks to all the 50+ participants — a mix of developers, data scientists, domain experts. The team spirit, work ethic and innovative results were amazing!
It was great to partner with the mVAM team from UN World Food Programme (WFP), with the generous support of our sponsors GoDataDriven, i3, and Coney. Want to know what challenges WFP brought to the table and results the different teams developed in just one day? Read on…
World Food Programme’s data driven challenges
WFP’s mobile Vulnerability Analysis and Mapping (mVAM) team focuses on mobile technology — for example SMS or interviews/surveys by phone — to collect information on food security in countries such as Sierra Leone and Yemen. This video shows how the mVAM team operates and collects survey data, to assist WFP in its mission for food security. Our hackathon results will help WFP make better use of its data. Two challenges were presented.
- Better data resolutions
The first challenge focused on data analysis. WFP wanted to improve the resolution of its data on food security estimates without having to collect more data, as this is expensive and time-consuming. WFP managers would benefit from more granular information, to identify areas with high food insecurity and design effective programmes to assist the most food insecure populations. Our hackathon teams came up with innovative solutions like neural networks trained through machine learning, and an investigation of the bias and systematic errors in survey data.
- Data visualization
The second challenge was all about data visualisation. WFP currently works with an online dashboard, but the mVAM component has yet to be developed. Our DataMissionaries’ challenge was to develop a decision-making tool for WFP managers that showcases clear data visualisation of the mVAM’s survey data and key messages on food security. Different dashboards and one app were developed. Some focused solely on mVAM’s survey data with the aim to present the insights as clearly as possible to the managers, while others took it a step further and tried to take into account other data like food prices and the influence of events like rainfall, conflicts and climate change.
A total of 50+ participants in ten teams worked for six hours, and, at the end of the day, pitched their results to the judges (3 mVAM team members, i3 sponsor and Centre4Innovation). Selecting the winner was a tough call, but eventually it came down to ideas that WFP felt most confident about developing further. The winners were:
- The overall winner was team 9 with Xiaowei Ma, Bouke Pieter Ottow, Andrew Ridden-Harper, Dennis van den Berg & Marijn van Zelst. They focused on challenge 1 and applied machine learning to create a neural network for Yemen. The neural network, when trained further, should be able to predict the Food Consumption Scores in unknown areas and can help WFP gain insight on better and more actionable food distribution.
- Second place was earned by Team 4 with Christel Veefkind-Gous, Jan-Joost van Kan, Michel Metselaar, Pieter Veefkind & Wouter Schuur. This mixed team with geo/data-analysts and developer delivered a geo-based dashboard for WFP managers (challenge 2) based on Sierra Leone data. The dashboard enables managers to get a first glance on important variables such as Food Security Score and Reduced Coping Strategy Index. Also, it is possible to drill down further with food prices and additional decision factors such as infrastructure, agricultural results, social stability, environmental disasters and diseases and epidemics.
- Team 2 won third prize with their approach to challenge 1. The team consists of Jiddu Alexander Broersma, Pablo Gabriel Celayas and Adriana Homolova. They focused on a solution to predict situations on district level. The starting point was district data that they felt most confident about and tried to apply machine learning models to train and model it for market data. This team defined very clearly the issues they encountered while working with the data.
The overall winner was awarded EUR100 to be donated to a charity of their choice; and all three winning teams received WFP gifts.
Let’s take it a step further…
WFP’s mVAM team was amazed by the quality of the solutions all teams delivered in just one day. Not only were tangible results and prototypes developed, WFP was also inspired by participants’ critical questions to get to the core of the challenges. The hackathon provided WFP with an opportunity to rethink and reflect on its way of working.
WFP is keen to further develop some ideas from the hackathon to fit its organisation: we will organise a follow-up event on Monday evening, June 13 in Amsterdam. More information and registration can be found on Eventbrite. We hope to see you there!
On behalf of DataMission, thank you all for the great participation at our first ever hackathon. We look forward to welcoming you to similar events going forward!
Blogs by participants
Want to hear about other participants’ experiences with DataMission’s hackathon? Read the blogs by Pieter Veefkind, Marijn van Zelst (Trifork) and Ice Mobile team. Also let us know your feedback, or if you have suggestions for future events! [:]
2 thoughts on “Solve hunger through data analysis and visualisation”