Authors:
Alexandros Bousdekis
1
;
Nikos Papageorgiou
1
;
Babis Magoutas
1
;
Dimitris Apostolou
2
and
Gregoris Mentzas
1
Affiliations:
1
National Technical University of Athens, Greece
;
2
University of Piraeus, Greece
Keyword(s):
Kalman Filter, Curve Fitting, Decision Support, Machine Learning, Manufacturing, Feedback.
Related
Ontology
Subjects/Areas/Topics:
Agents
;
Applications of Expert Systems
;
Artificial Intelligence
;
Artificial Intelligence and Decision Support Systems
;
Enterprise Information Systems
;
Industrial Applications of Artificial Intelligence
;
Intelligent Agents
;
Internet Technology
;
Operational Research
;
Web Information Systems and Technologies
Abstract:
The incorporation of feedback in the proactive event-driven decision making can improve the recommendations generated and be used to inform users online about the impact of the recommended action following its implementation. We propose an approach for learning cost functions from Sensor-Enabled Feedback (SEF) for the continuous improvement of proactive event-driven decision making. We suggest using Kalman Filter, dynamic Curve Fitting and Extrapolation to update online (i.e. during action implementation) cost functions of actions, with the aim to improve the parameters taken into account for generating recommendations and thus, the recommendations themselves. We implemented our approach in a real proactive manufacturing scenario and we conducted extensive experiments in order to validate its effectiveness.