Towards privacy-preserving digital marketing: an integrated framework for user modeling using deep learning on a data monetization platform
Qiwei Han, Carolina Lucas, Emila Aguiar, Patrícia Macedo, Zhenze Wu
Abstract
Abstract This paper presents a novel approach to privacy-preserving user modeling for digital marketing campaigns using deep learning techniques on a data monetization platform, which enables users to maintain control over their personal data while allowing marketers to identify suitable target audiences for their campaigns. The system comprises of several stages, starting with the use of representation learning on hyperbolic space to capture the latent user interests across multiple data sources with hierarchical structures. Next, Generative Adversarial Networks are employed to generate synthetic user interests from these embeddings. To ensure the privacy of user data, a Federated Learning technique is implemented for decentralized user modeling training, without sharing data with marketers. Lastly, a targeting strategy based on recommendation system is constructed to leverage the learned user interests for identifying the optimal target audience for digital marketing campaigns. Overall, the proposed approach provides a comprehensive solution for privacy-preserving user modeling for digital marketing.