Chapter 8 Unsupervised Learning: An Application in International Trade

Description

The field of international trade has evolved significantly with the rise in importance of analyses at levels of finer granularity than traditional macroeconomic approaches. In this context, new data are needed to inform new theories. In this session, we discuss the fundamental concepts of unsupervised learning. In particular, we will look at the usefulness of these approaches for international trade data. In this context, we will choose a particular application: to address the issue of regionalization versus globalization created by bilateral trade agreements that complement the World Trade Organization. Why this question? Without the tools of data science, the answer to this question would indeed be limited in terms of methodology. In this session, we will examine the United Nations International Trade and Development project using data science methods.

Concepts discussed :

• 1 K-means

• 2 silhouette coefficient

• 3 collaborative filtering

• 4 methods of the nearest K neighbors

• 5 cosine similarity

Pre-Session Activities/Resources

• United Nations Conference on Trade and Development “Text-as-Data analysis of Trade Agreements” (http://mappinginvestmenttreaties.com/rta/) Activities/Resources during the session

• Alschner, Wolfgang, Julia Seiermann, and Dimitriy Skougarevskiy. 2017. “Text-as-Data Analysis of Preferential Trade Agreements: Mapping the PTA Landscape.” 2017. https://unctad.org/en/pages/PublicationWebflyer.aspx?publicationid=1838.

Post-Session Activities/Resources

General Resources

• gapminder by Jenny Bryan

• Leaflet for R

• Flexdashboard for R

• Leaflet Cheat Sheet

Weblinks

https://towardsdatascience.com/t-sne-clearly-explained-d84c537f53a