แนวคิดหลัก
Adapting genetic algorithms for counterfactual generation in Predictive Process Monitoring with temporal constraints.
บทคัดย่อ
The article discusses the importance of generating counterfactual explanations in Predictive Process Monitoring, focusing on maintaining control flow relationships. It introduces adaptations to genetic algorithms to incorporate temporal background knowledge, ensuring feasibility and adherence to process constraints. The study evaluates these methods against state-of-the-art techniques using real-life datasets.
สถิติ
"State-of-the-art efforts in PPM have focused on delivering accurate predictive models through the application of ensemble learning and deep learning techniques."
"Counterfactual explanations suggest what should be different in the input instance to change the outcome of an AI system."
"The proposed methods are evaluated with respect to state-of-the-art genetic algorithms for counterfactual generation."
คำพูด
"Counterfactual explanations are essential for providing alternatives to achieve a certain outcome in the PPM domain."
"None of the previous approaches make use of background knowledge explicitly when generating counterfactual explanations."