Machine Learning as a Service (MLaaS) Selection for IoT Environments

Abstract

This thesis presents two novel frameworks for selecting Machine Learning as a Service (MLaaS) providers using incomplete Quality of Service (QoS) information and contextual data in IoT environments. The proposed MLaaS Selection Framework (MSF) enhances service selection with bias detection and explainability mechanisms, while the IoT-based framework dynamically maps user contexts to MLaaS services. Together, these frameworks improve service efficiency, accuracy, and responsiveness, enabling informed MLaaS selection based on user preferences and contextual changes.