The complex and opportunistic environment in which edge computing systems operate, poses a fundamental challenge for online edge system orchestration, resource provisioning and real-time responsiveness in response to user movement. Such a challenge needs to addressed throughout the edge system lifecycle, starting from the software development methodologies. In this paper, we propose a novel development process for modeling opportunistic edge computing services, which rely on (i) ETSI MEC reference architecture and Opportunistic Internet of Things Service modeling for the early stage of system analysis and design, i.e. domain model and service metamodel; and on (ii) feature engineering for evaluating those opportunistic aspects with data analysis. To address the identified opportunistic properties, at the service design phase we construct (both automatically and through domain expertise) Opportunistic Feature Vectors for Edge, containing the numerical representations of those properties. Such vectors enable further data analysis and machine learning techniques in the development of distributed, effective and efficient edge computing systems. Lastly, we exemplify the integrated process with a microservice-based user mobility management service, based on a real-world data set, for online analysis in MEC systems.