XGBoost and Random Forest Based SME Export Readiness Prediction for African Digital Entrepreneurship Ecosystem Development

Small and medium-sized enterprises are central to African economic diversification, yet many digitally enabled SMEs lack reliable decision-support systems for assessing export readiness before entering regional and international markets. This paper proposes a technical machine learning framework for predicting SME export readiness within African digital entrepreneurship ecosystems using a novel hybrid ensemble model named ExportReadiness-XRFNet, which integrates XGBoost, Random Forest, feature-weighted stacking, and probability calibration. The model is designed to classify SMEs into low, moderate, and high export-readiness categories using structured indicators such as digital capability, financial stability, product standardization, logistics preparedness, regulatory compliance, market intelligence, e-commerce adoption, and cross-border payment readiness. The proposed model is compared with Logistic Regression, Support Vector Machine, Decision Tree, Gradient Boosting, standalone Random Forest, and standalone XGBoost models. Graph-based performance analysis, including accuracy curves, ROC-AUC plots, confusion matrices, feature-importance rankings, and precision-recall comparisons, is used to evaluate predictive superiority. The expected results demonstrate that the hybrid ExportReadiness-XRFNet model achieves stronger classification accuracy, improved recall for export-ready SMEs, better handling of nonlinear feature interactions, and more interpretable readiness drivers than conventional baseline models. The study contributes a scalable AI-based decision-support framework for policymakers, SME development agencies, export-promotion councils, fintech platforms, and digital entrepreneurship hubs seeking to identify, support, and scale African SMEs with high export potential.