Model Validation
摘要
In this chapter, we specify the business requirements and propose the solution concept for model validation. Statistical methods exist to validate artificial intelligence models regarding the accuracy of predictions. However, for ERP business applications, this mathematical approach alone is not enough. Additional validation from functional and business process points of view are necessary. Existing ERP techniques like A/B testing, feature toggle, business features, or switch framework must be analyzed in this context. We suggest a validation process that is appropriated for ERP software and considers the state transitions of artificial intelligence models. Finally, we propose a model validation cockpit as central environment for administration.