Context-aware IoT Service Recommendation: A Deep Collaborative Filtering-based Approach

Abstract

The advantages of Service-Oriented Architecture (SOA) combined with the emerging and diffusion of the Internet of Things (IoT) instances have given birth to a new paradigm for IoT components integration, i.e., Service-Oriented IoT. Microservices especially have been widely used to deliver IoT services due to their lightweight implementation and distributed nature. With the continuous increase in IoT services available on the Internet, the selection of services becomes difficult. Furthermore, IoT services are often featured with rich contexts and OpenAPI descriptions, which impede service recommendation approaches that are designed for WSDL-based Web services or mashup services. To address this challenging issue, we propose a context-aware IoT service recommendation approach called DFORM for proactive service provision. DFORM considers both functional features of OpenAPI descriptions and contextual features of IoT environments, and leverages a deep collaborative filtering-based recommendation model to learn the feature representations and capture the interactions between users and services. We conduct a series of experiments to evaluate the recommendation performance of DFORM and the experimental results show that DFORM is effective in IoT service recommendation and outperforms state-of-the-art techniques.