The rapid development and implementation of smart, connected products (SCPs) in the engineering field has triggered a promising manufacturing paradigm of servitization, i.e. smart product-service systems (Smart PSS). As a complex solution bundle in both system and product level, its engineering change management differs from the existing ones mainly in two aspects. Firstly, massive in-context stakeholder-generated/product-sensed data during usage stage can be leveraged to enable its success in a data-driven manner. Secondly, the digitalized services, consisting of both hardware and software solutions, can also be changed in a more flexible way other than the physical components alone. Nevertheless, scarcely any work reports on how to conduct engineering change in such context, let alone a systematic approach to support the automatic generation of its change prediction or recommendation. Aiming to fill these gaps, this work proposes an occurrence-based design structure matrix (DSM) approach together with a three-way based cost-sensitive learning approach for automatic engineering change management in the Smart PSS environment. This informatics-based research, as an explorative study, overcomes the subjectivity and tedious assessment of the experts in the conventional approaches, and can offer useful guidelines to the manufacturing companies for managing their engineering changes for product-service innovation process.