New article published in “Nature Sustainability” on enabling a global monitoring of development aid with machine learning

Through the global development aid system, more than 200 billion US-Dollars are spent annually, supporting various activities in developing countries, such as building schools, electrifying villages, or facilitating access to health care. As existing approaches to monitor development projects are based on manual reporting, they are highly bureaucratic, expensive and, in many areas, fail to provide a comprehensive and up-to-date overview.

In a new study, Malte Toetzke, Nicolas Banholzer, and Stefan Feuerriegel have developed a machine learning framework that categorizes millions of development projects based on their textual descriptions. Based on the framework, they clustered 3.2 million project descriptions into 173 “activity clusters” and thereby enabled a new, comprehensive, and granular categorization of global development projects between 2000 and 2019. This allowed them to monitor what kind of development projects have been financed globally over the past decades—capturing many topics that have not yet been analyzed systematically, such as conservation of forests, youth empowerment, or microfinance. Furthermore, the monitoring reveals funding gaps across recipient countries as well as new trends in the topics of global aid activities.

The presented framework can contribute to an improved and more timely coordination of global aid spending between donors. Furthermore, the data can spark new research investigating the relation between global development aid and progress towards the Sustainable Development Goals. Code and Data can be found here: Downloadhttps://github.com/MalteToetzke/Monitoring-Global-Development-Aid-With-Machine-Learningvertical_align_bottom.
Full text access: Downloadhttps://www.nature.com/articles/s41893-022-00874-zvertical_align_bottom