Intelligent monitoring systems are gaining more and more relevance with many possible applications, from industrial settings to assistive devices for humans. For many possible functions, prediction is one of the basic capabilities to provide the right information at the right time. Machine Learning models should be able to cope with concept drift in streaming non-stationary data. The aim of this project is to extend existing learning methods towards more heterogeneous and challenging learning scenarios. A particular focus of this project is to apply advanced Machine Learning research to real-world problems within the Honda Research Institute EU (HRI-EU) for example smart energy monitoring systems or models for exoskeleton control.