Model Degradation
摘要
In this chapter, we specify the business requirements and propose the solution concept for model degradation. Over the course of time, the prediction power of artificial intelligence models decreases due to changes in the data environment. Determining this time point and triggering retraining is the objective of model degradation. However, this is a challenging task as in addition to statistical techniques, feedback of the users is required and also the ability to parallelly run models in ERP systems. Our focus is less on the data science methods for model degradation but on the integration aspects regarding ERP software.