A data-driven framework for assessing global progress towards sustainable development goals

Abstract

Effective monitoring of the Sustainable Development Goals (SDGs) is crucial for advancing global sustainable development. However, widespread data gaps continue to hinder the accurate assessment of SDG performance across countries and goals. To address this challenge, this study develops a data-driven integrated assessment framework combining dimensionality reduction and machine learning-based imputation techniques, based on 380 SDG indicators from the World Bank database covering the period 2000–2020. Principal indicators were selected using a combination of Principal Component Analysis (PCA) and multiple regression, and missing data were imputed using the random forest (RF)-based missForest algorithm. Based on the completed dataset, the SDG index and performance of 17 individual SDGs were assessed for 215 countries and regions worldwide from 2000 to 2020. The results show that: (1) identification of 218 principal indicators covering over 90 % of the information in the initial set; (2) robust imputation of missing values with a Normalized Root Mean Squared Error (NRMSE) of approximately 0.2 and a Proportion of Falsely Classified (PFC) around 0.08; (3) a steady global improvement in SDG performance with significant regional disparities—Europe leading, Africa lagging, and Asia progressing most rapidly; and (4) uneven development across different goals, with some facing considerable challenges. This study enhances the completeness and applicability of global SDG performance assessment and provides empirical evidence to support more targeted sustainable development policymaking.