Replication Data and Code for “Structural Transformation and Export Competitiveness in Africa and Asia: Evidence from Interpretable Machine Learning”

This dataset supports the study titled “Structural Transformation and Export Competitiveness in Africa and Asia: Evidence from Interpretable Machine Learning.” The study examines how sectoral transformation, labour reallocation, digital readiness, infrastructure, foreign direct investment, income capacity, and labour-force scale predict export competitiveness across African and Asian economies. The repository contains the processed analytical dataset and Python code used for the empirical analysis. The data cover 123 economies from Africa and Asia over the period 1985 to 2024, producing a balanced country-year panel of 4,920 observations. The outcome variable is exports of goods and services as a percentage of GDP, used as the export competitiveness benchmark. The modelling predictors include industry value added, services value added, agriculture value added, employment in industry, employment in services, employment in agriculture, internet use, access to electricity, foreign direct investment inflows, GDP per capita, and total labour force. Country, continent, subregion, and year are retained for descriptive interpretation but excluded from the supervised modelling features. The accompanying Python notebook reproduces the full analysis. It compares linear regression, ridge regression, random forest, and gradient boosting models using train-test validation and five-fold cross-validation. Model performance is assessed with R-squared, mean absolute error, and root mean squared error. The notebook also generates permutation feature importance for the best-performing model, principal component analysis, and K-means clustering to identify structural export-profile groups. The files are shared to support transparency, replication, and further research on structural transformation, labour reallocation, export competitiveness, and machine learning applications in development economics. The data were drawn from public development indicators and processed for the purpose of this study.