Adaptive preference transfer for personalized IoT entity recommendation

Abstract

• An IoT entity recommendation method based on adaptive preference transfer is proposed. • A lightweight user feedback feature embedding method is proposed to solve the problem of sparse user feedback. • The rigorous experiments are designed to verify the validity of the proposed method. Internet of Things (IoT) recommendation can effectively improve the convenience and intelligence of users in obtaining information about interested physical entities in a huge entity search space. However, existing transfer learning-based recommendation methods mainly focus on Internet resources and services, lack personalized transfer weight consideration for different user preferences, and fail to address the entity rating matrix sparsity problem, resulting in limited IoT entity recommendation performance. Thus, we design an adaptive preference transfer IoT entity (APTE) recommendation method in this paper. Firstly, considering various user preference characteristics, an adaptive dual-domain transfer model for item domain and social domain is designed to meet users’ personalized requirements. Then, a lightweight user feedback embedding method is proposed to mine the explicit and implicit features and to embed the entity rating matrix to solve the problem of sparse user feedback information. Simulation results demonstrate that APTE can effectively improve the recommendation performance compared with state-of-the-art baselines.