Pervasive applications and services are increasingly based on the intelligent interpretation of data gathered via heterogeneous sensors dipped in the environment. Classical Machine Learning (ML) techniques often do not go beyond a basic classification, lacking a meaningful representation of the detected events. This paper introduces a early proposal for a semantic-enhanced machine learning analysis on data of sensors streams, performing better even on resource-constrained pervasive smart objects.
The framework merges an ontology-driven characterization of statistical data distributions with non-standard matchmaking services, enabling a fine-grained event detection by treating the typical classification problem of ML as a resource discovery.