Introducing Data Science Techniques for Trade Analysis: With Applications in Knime
‘Data science’ refers to a growing body of knowledge around the acquisition, manipulation, visualisation, analysis and reporting; usually by means of analysis tools or platforms. The ultimate goal of data science practice is to be able to predict behaviour such that some form of remedial or reactive action can be taken – whether for profit, policy or other purpose.
This Working Paper explores the basics of statistical learning, as applied to a trade analysis problem and using the application known as Knime Desktop Analytics Platform. An attempt is made to develop a machine learning method to predict whether an African country will be ‘converging’ (as in, increasing its relative intra-African trade) or not. The paper details the steps required to prepare the data using ETL (extract-transfer-load) techniques, before describing the process to prepare and run a machine learning algorithm.
The form of classification utilised here is non-parametric, but an attempt is made to compare the approach to the usual parametric one utilised by econometricians. Parametric methods explicitly define an objective variable and the determinants of that variable, and attempt to explain and predict.
The results indicate that data science techniques are indeed powerful and useful to trade analysts, but that more work is needed to build a more accurate predictor of intra-African convergence.
Readers are encouraged to quote and reproduce this material for educational, non-profit purposes, provided the source is acknowledged. All views and opinions expressed remain solely those of the author and do not purport to reflect the views of