Performance Evaluation of Sustainable Development Goals Employing Unsupervised Machine Learning Approach

Abstract

NITI Aayog has been publishing different reports on the performances of various Indian States and union territories (UTs) under the Sustainable Development Goals (SDGs) agendas. For this purpose, various socio-economic information related to poverty, food security, health care, employment, terrestrial ecosystems, law and order, etc., has been used. NITI Aayog has applied different statistical transformations such as normalization on these data points and applied a straightforward, globally accepted, and robust classification methodology. The objective of NITI Aayog’s methodology is to generate an aggregated score for every States/UTs based on the achievement of the respective goals keeping the national level target as optimal level to adhere. Finally, NITI Aayog has classified the States/UTs into different clusters based on the aggregated scores. In this paper, we have considered the growth rate of above-stated data points to capture year-on-year progression of States/UTs on each SDGs and applied machine learning-based classification algorithms to create different homogeneous clusters of States/UTs. We have analysed the characteristics of each cluster and tried to identify the important differentiating factors. We also compare the results of different machine learning algorithms and find out the similarity of the solutions produced by these algorithms.