A Tool Framework for KPI Application Development

This paper describes a KPI (key performance indicator) modeling environment, coined Mozart, where modelers can use formal models to explicitly define the services of KPI and their relationships which are depicted by KPI net. Mozart provides us with methods for mining and modeling KPIs and supports smooth model transformation for generating monitoring applications based on a model driven approach. It also provides us with methods of service composition for KPI applications. We showed how it works with an example scenario of automobile data and found that "mpg" is most strongly influenced by four KPIs. The result showed which KPIs should be focused for human monitoring, and it can be an initial model for monitoring applications.