Using Machine Learning to map impact evaluations worldwide
New technologies can contribute to bridging the evidence gap by identifying which interventions have already been assessed through impact evaluations. Using machine learning, the graph below shows the results of an analysis of about 40 million open access scholarly publications since early 2000. Only around approximately 37,000 impact evaluations were identified using a hybrid approach with I) relevant keyword search in abstracts and ii) a trained topic model for a subset of 15,000 studies from organizations such as the World Bank, J-PAL, 3ie, Oxfam and WFP. Therefore, despite the raising trend in the number of impact evaluations published worldwide, the estimated amount of available studies is still limited.
Coming Soon
Mapping impact evaluations of interventions around UNDP’s signature solutions – poverty and inequality, gender, environment, energy, governance and resilience.
