Research on the Construction of Differentiated Pricing Model for SMEs’ Data Assets Based on Machine Learning in the Context of Digital Economy

This study focuses on the small sample size and high heterogeneity of data assets owned by small and medium-sized enterprises in the digital economy. Through questionnaire surveys and factor analysis, four core pricing factors are identified, including data characteristics, enterprise heterogeneity, market environment, and technology application. A five-stage pricing framework consisting of data augmentation, feature selection, model training, pricing output, and dynamic adjustment is then constructed. The model integrates CTGAN-based data augmentation and the XGBoost algorithm, and embeds an enterprise-heterogeneity adaptation layer to achieve differentiated pricing. Empirical results show that the proposed model performs well, with MAE of 0.072, RMSE of 0.095, and R2 of 0.83. The pricing error rate is below 10%, significantly outperforming traditional methods and undifferentiated machine-learning benchmarks. The model also demonstrates good effectiveness and stability. This research provides operable models and tools for accurate pricing of SMEs’ data assets and offers a technical reference for promoting standardized development of the data-factor market.