![]() ![]() Although higher frequency surveys can improve the timeliness of survey coverage, this often comes at the cost of spatial and survey detail. Moreover, high-quality, large-scale surveys are typically fielded only once every several years, and are generally statistically representative of the population only at relatively large (e.g., provincial or regional) scales under standard sampling protocols. Such surveys are, however, expensive and time-consuming, and may systematically omit subregions that are harder or more dangerous to physically access, despite the severe poverty and malnutrition prevalence often endemic to such locations. ![]() To inform aid targeting, monitoring and evaluation efforts, agencies have historically drawn data mainly from detailed household surveys, such as the large-scale, nationally-representative Demographic and Healthy Surveys (DHS) or Living Standards Measurement Study (LSMS) programs. The more precise and interpretable the estimates, and the more parsimonious and inexpensive the data demands of the model, the greater the likelihood that agencies can employ such methods to accurately target and evaluate interventions to address agricultural, economic, political, or weather shocks that might otherwise thrust vulnerable groups into poverty traps or famine. Contemporaneous prediction-i.e., estimation of locations not covered in standard household surveys-can fill in the gaps in survey evidence, generating poverty maps for geographic targeting of interventions and to inform ongoing monitoring and evaluation activities. Nowcasting-i.e., using current observations of predictive features combined with past observations of the poverty or malnutrition outcome(s) of interest-can help with geographic needs assessments and targeting, as well as provide baseline measures for impact evaluation of interventions. Accurate identification of poor or malnourished populations in space and time serves multiple purposes. One key factor in the effectiveness of such efforts is the accuracy with which poor and malnourished populations can be identified. Governments and humanitarian agencies devote considerable resources towards poverty and malnutrition reduction efforts. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.Ĭompeting interests: The authors have declared that no competing interests exist. USAID website: Funding was received by CB (Chris Barrett, PI), YS, LH, YL, DM (co-pis). This data has also been attached as supplemental information accompanying this manuscript.įunding: This work was funded by the United States Agency for International Development under cooperative agreement # 7200AA18CA00014, “Innovations in Feed the Future Monitoring and Evaluation - Harnessing Big Data and Machine Learning to Feed the Future”. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.ĭata Availability: Links to all data sources considered in this work can be found in Table S5, while the processed data used in this work is available for download at. Received: MaAccepted: JPublished: September 8, 2021Ĭopyright: © 2021 Browne et al. United Nations University Institute for Natural Resources in Africa, GHANA (2021) Multivariate random forest prediction of poverty and malnutrition prevalence. Applied to data from 11 low and lower-middle income countries, we find predictive accuracy broadly comparable for both tasks to prior studies that use proprietary data and/or deep or transfer learning methods.Ĭitation: Browne C, Matteson DS, McBride L, Hu L, Liu Y, Sun Y, et al. We demonstrate two use cases: contemporaneous prediction, which might be used for poverty mapping, geographic targeting, or monitoring and evaluation tasks, and a sequential nowcasting task that can inform early warning systems. We demonstrate how interpretable random forest models can produce estimates of a set of (potentially correlated) malnutrition and poverty prevalence measures using free, open access, regularly updated, georeferenced data. However, state of the art models often rely on proprietary data and/or deep or transfer learning methods whose underlying mechanics may be challenging to interpret. Advances in remote sensing and machine learning enable increasingly accurate, inexpensive, and timely estimation of poverty and malnutrition indicators to guide development and humanitarian agencies’ programming. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |