An Interpretable Semi-Parametric Framework for CMM Export Prediction via Symbolic Regression and Multi-Objective Optimization

Amid the accelerating globalization of Traditional Chinese Medicine (TCM), the export performance of Chinese Medicinal Materials (CMMs) has emerged as a critical indicator of its international competitiveness and systemic resilience. However, existing studies predominantly rely on aggregated national statistics, thereby obscuring substantial inter-provincial heterogeneity and limiting the capacity to capture underlying structural dynamics. To address this challenge, this study conceptualizes CMM export as a complex regional economic system and constructs a panel dataset encompassing 31 provinces in China. We propose a structurally interpretable semi-parametric framework, termed the SR-MOABC-FE model, which integrates symbolic regression (SR) with a fixed effects (FE) specification. Within this framework, SR is employed to flexibly characterize nonlinear relationships, while the FE component accounts for province-specific heterogeneity. To further enhance model performance, a multi-objective artificial bee colony (MOABC) algorithm is developed to simultaneously optimize goodness-of-fit, predictive accuracy, and model parsimony. Empirical results demonstrate that the proposed model achieves a test-set mean absolute percentage error (MAPE) of 9.15%, significantly outperforming both conventional econometric models and mainstream machine learning approaches. Beyond predictive gains, the model retains strong structural interpretability, enabling the identification of key driving factors, including policy support, rural residents’ per capita disposable income, and natural resource endowment. Overall, this study advances a hybrid system modeling paradigm that bridges interpretable econometric structures and data-driven symbolic learning, offering both methodological innovation and actionable insights for the analysis and optimization of complex trade systems.