Model-Based Recommendations for Next-Best Actions in Knowledge-Intensive Processes

Knowledge-intensive processes are highly flexible and volatile and therefore hard to predict. During the execution of a case, knowledge workers plan their actions to reach their goals based on their expertise and traditionally without system support. This paper proposes a model-driven framework that allows combining process model analysis with existing predictive process monitoring approaches to automatically recommend the next best action to perform. The approach is based on case-specific and data-centric goals that the knowledge workers may specify during run-time. A prototypical implementation uses state space analysis based on colored Petri nets to compute recommendations, while user experiments show their value to knowledge workers.