Concept for a Technical Infrastructure for Management of Predictive Models in Industrial Applications

Abstract

With the increasing number of created and deployed prediction models and the complexity of machine learning workflows we require so called model management systems to support data scientists in their tasks. In this work we describe our technological concept for such a model management system. This concept includes versioned storage of data, support for different machine learning algorithms, fine tuning of models, subsequent deployment of models and monitoring of model performance after deployment. We describe this concept with a close focus on model lifecycle requirements stemming from our industry application cases, but generalize key features that are relevant for all applications of machine learning.

Continuous improvement and adaptation of predictive models in smart manufacturing and model management

“…Kriens et al [17] propose a definition of model metadata to describe model capabilities and software dependencies to allow provisioning of execution environments for models. Recent efforts towards model management systems for smart manufacturing [18][19][20][21][22] try to apply these findings in an industrial setting. Weber et al [20] state that current model management systems, when applied in smart manufacturing environments, lack the integration of external knowledge bases to provide domain context and lack proper mechanisms to retire models at the end of their lifetime.…”

“…Recent efforts towards model management systems for smart manufacturing [18–22] try to apply these findings in an industrial setting. Weber et al.…”

“…We presented our own concept for an infrastructure for management of predictive models in [18]. In this concept we focus on the application of predictive models in smart manufacturing scenarios with its continuous adaptation of models to changing context's.…”