To address challenges such as diversified user demands and potential data privacy risks on cross-border e-commerce platforms, this study proposes an improved Siamese network model tailored for multi-faceted user needs. The model comprises three modules. First, a data-driven knowledge structuring module integrates multimodal features. Second, a privacy-preserving contrastive learning module with federated optimization enables distributed training while safeguarding user data privacy. Third, a multi-objective optimization module incorporates diverse user requirements, producing outputs that simultaneously consider multiple performance and business objectives. Experimental results demonstrate that the proposed model outperforms baseline methods, achieving an F1-score of 0.841 and an Area Under the Curve (AUC) of 0.923. In terms of business-relevant metrics, the model also achieves superior performance, with a diversity index of 0.175 and a fairness disparity of 0.052. These results indicate that the proposed model effectively balances multiple user demands and privacy protection in cross-border e-commerce contexts, providing a reference framework for platform system optimization.
Hou et al. (Thu,) studied this question.