Application of Machine Learning in Forecasting International Trade Trends

Feras A. Batarseh Graduate School of Arts & Sciences - Data Analytics Georgetown University Washington, D.C. 20057 feras.batarseh@ georgetown.edu

Munisamy Gopinath
Department of Agricultural and Applied Economics University of Georgia Athens, Georgia 30602 m.gopinath@uga.edu

Ganesh Nalluru Volgenau School of Engi- neering George Mason University Fairfax, Virginia 22030 gn@gmu.edu

Jayson Beckman Economic Research Services
Department of Agriculture Washington, D.C. 20024 jayson.beckman@usda.gov

Abstract – International trade policies have recently gar- nered attention for limiting cross-border exchange of es- sential goods (e.g. steel, aluminum, soybeans, and beef). Since trade critically affects employment and wages, pre- dicting future patterns of trade is a high-priority for policy makers around the world. While traditional economic models aim to be reliable predictors, we consider the pos- sibility that Machine Learning (ML) techniques allow for better predictions to inform policy decisions. Open- government data provide the fuel to power the algorithms that can explain and forecast trade flows to inform policies. Data collected in this article describe international trade transactions and commonly associated economic factors. Machine learning (ML) models deployed include: ARIMA, GBoosting, XGBoosting, and LightGBM for predicting future trade patterns, and K-Means clustering of countries according to economic factors. Unlike short-term and sub- jective (straight-line) projections and medium-term (aggre- gated) projections, ML methods provide a range of data- driven and interpretable projections for individual com- modities. Models, their results, and policies are introduced and evaluated for prediction quality.

Keywords: Machine Learning, International Trade, Boosting, Predictions, Imports and Exports Motivation and Background
In recent years, many countries are concerned about rising trade deficits (value of exports less imports) and their im- plications for employment and wages. For instance, the United States’ goods and services trade deficit with China was $378.8 billion in 2018. Such numbers are forcing

countries to either exit trade agreements or enforce tariffs, (e.g. Brexit, U.S. tariffs on Chinese goods). These shocks to global trade in commodities pose challenges to predict future trading patterns. In the United States, farm program costs, the President’s Budget and recent compensation pro- grams to address farmers’ losses due to retaliatory tariffs depend on accurate trade predictions (from: USDA, 2018 – Trade damage estimation).
International economics has a long history of improving our understanding of factors causing trade and the conse- quences of free flow of goods and services across coun- tries. Nonetheless, the recent shocks to the free-trade re- gime raise questions on the quality of earlier predictions and their applicability in the context of large trade disputes (Batarseh et al. 2018). To address these challenges, this article, identifies ML techniques appropriate for the inter- national trade setting and tests their validity in making high quality projections. Recent technological advancements in ML as well as data democratization have also helped transparency, which is critical in the context of trade poli- cy-making. Given the Open Data and Big Data initiatives presented in 2008 and 2012 (White House 2008), federal agencies are forced to share their data on public reposito- ries such as www.data.gov . Econometric approaches iden- tified have multiple ingredients that effect commodities’ production and utilization, and hence, directly influence imports and exports of those commodities (Gevel et al. 2013). A field that can greatly aid with this analysis is Ex- plainable AI (XAI) (Gunning 2019). XAI and ML can be instrumental in explaining previous and emerging patterns in data. This paper aims to address three main questions: 1- Do economic variables (such as GDP and population) associate with each other and a countries’ exports? 2- Can boosting algorithms ensure learning and predictions from country-commodity-year cubical trade data? 2

3- Can ML techniques qualitatively improve the forecast from traditional econometrics? We answer these questions in this article using experi- mental work. We also find that the ML data models devel- oped here are scalable to all trade transactions, all over the world, and for all commodities. Related Work Based on a recent study by the National Bureau of Eco- nomic Research (NBER), ML is only recently being ap- plied to econometrics. ML has been applied across multi- ple domains; it has been employed in addressing challeng- es in healthcare (Reddy and Aggarwal 2015), education (Niemi et al.), and sports (Alamar 2013). To d

…(Full text truncated)…