eduzhai > Applied Sciences > Computer Science >

A new set of cluster driven composite development indicators

  • KanKan
  • (0) Download
  • 20210425
  • Save

... pages left unread,continue reading

Document pages: 26 pages

Abstract: Composite development indicators used in policy making often subjectivelyaggregate a restricted set of indicators. We show, using dimensionalityreduction techniques, including Principal Component Analysis (PCA) and for thefirst time information filtering and hierarchical clustering, that thesecomposite indicators miss key information on the relationship between differentindicators. In particular, the grouping of indicators via topics is notreflected in the data at a global and local level. We overcome these issues byusing the clustering of indicators to build a new set of cluster drivencomposite development indicators that are objective, data driven, comparablebetween countries, and retain interpretabilty. We discuss their consequences oninforming policy makers about country development, comparing them with the topPageRank indicators as a benchmark. Finally, we demonstrate that our new set ofcomposite development indicators outperforms the benchmark on a datasetreconstruction task.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×