IoT service platform enhancement through ‘in-situ’ machine learning of real-world knowledge

With Machine-to-Machine and Internet of Things getting beyond hype, including an ever wider range of connected device types in ever more value-added services, a new era of data (and multimedia) stream-intensive services is emerging. While live data is massively becoming available, turning it into meaningful information that is not only actionable for decision makers, but also can be leveraged as a behavioral service property, or even reused across services, is a challenge that demands a systematic approach. In this paper we propose such systematic approach, towards establishing an Internet of Things service platform architecture that leverages real-world knowledge for faster service creation and more efficient execution. Illustrated by example scenarios, we go further beyond this, proposing a method to systematically leverage machine learning techniques for revising, improving or ultimately semi-automatically extending this real-world knowledge `in-situ´, i.e. during system operation, leveraging real-world observation in-context of requested service execution.