Privacy-preserving K-means for Customer Segmentation in Precision Marketing

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Objective The existing customer segmentation approaches in precision marketing are faced with some problems, including the lack of coordination between the protection of customer privacy and the accuracy of customer segmentation, as well as the difficulty of adapting to the heterogeneity of the data sources. This study proposes a customer segmentation model based on the K-means Clustering method and the Federated Learning approach. Method The model uses K-means Clustering as the core and integrates the privacy-preserving features of Federated Learning. Clustering computation is performed locally at each data node, and the distributed clustering results are securely aggregated to produce accurate customer segmentation outcomes. Results From the experimental results, it can be observed that the proposed model has an accuracy of 98.76% in customer segmentation and a risk of 2.97% in privacy leakage. For the high-frequency consumption customer test scenario, the accuracy of segmentation compliance is 98.63%. For the targeted marketing scenario for maternal and infant products communities, the efficiency of training is 128.36 data/s and interpretability of segmentation is 0.946. Conclusion The results indicate that the proposed customer segmentation model significantly outperforms comparison methods in both segmentation accuracy and privacy protection. It effectively addresses issues such as cross-modal semantic inconsistency and delayed capture of dynamic preferences, providing a feasible technical solution for enhancing privacy security and segmentation quality in precision marketing.