Grant: Discovery Grants Program - Individual
The University of British Columbia
$43,000.00
Combined Statistical Learning and Robust Optimization
EF 2024-25
Résumé
Most optimization models in industrial engineering account for data with a statistical model, and decisions are made under the assumption the model is correct. On the other hand, models that account for possible model miss-specification, often formulated as robust optimization models, do not account for learning. Such models essentially assume that data is of no value when it comes to learning the model.
The proposed research program aims to develop new models in which a decision maker can account for possible model miss-specification while also learning at the same time. Three streams of research will be investigated.
1. Robust Maintenance Optimization with Learning. Maintenance optimization with learning and model miss-specification has not been considered in the literature. This stream will develop new models that account for learning from data and generate policies that work well even when the underlying model is flawed. The theory will be applied on real-world sensor data to investigate the performance of such models, and to develop visualization and recommendation tools.
2. Robust Inventory Control with Learning. Multi-stage inventory control models with learning and model miss-specification will be developed. While preliminary result shows that the myopic-optimal order quantity is "boosted" when accounting for learning and reduced when accounting for model miss-specification, more research is needed to better characterize these adjustments, and to conduct rigorous validation tests using datasets from Canadian partner companies.
3. Optimization with Multiple Biased Data. A class of multi-source Bayesian control problems wherein the decision-maker can acquire information from multiple, possibly biased sources before making a decision will be investigated. The model will find value in healthcare settings where decisions are based on combining data from different, possibly biased sources. For example, in medical diagnosis practice, it is common to use multiple screening tests or tools (e.g., X-ray, blood work, CT scan) to increase test accuracy and reliability.
If successful, the proposed research will not only provide practical value to companies but lead to new insights explaining the underlying mechanism by which one trades off between learning and model miss-specification. We will disseminate our findings through publications, conferences and practitioner--oriented industry councils. Doctoral students will be trained as future scholars, while those entering industry will champion data analytics in the private sector or the government. (HQPs account for a majority of the budget.) The broader impact will be the improvement of industry practice and educational programs. The research will be conducted in collaboration with Canadian companies connected with the UBC Sauder, bringing together students, faculty and companies to allow for testing of new ideas on datasets, as well as the direct dissemination of knowledge into classrooms.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
British Columbia, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2024-05213
Name nom
Kim, Michael
Department département
Sauder School of Business
Organization id
2
Institution établissement
The University of British Columbia
Province en
British Columbia
Province fr
Colombie-Britannique
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
43000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1509
Committee name en
Civil, Industrial and Systems Engineering
Committee name fr
Génie civil, industriel et des systèmes
Area of application code
520
Area of application group en
Manufacturing processes and products
Area of application group fr
Procédés et produits industriels
Area of application en
Production and operations management
Area of application fr
Production et gestion des opérations
Research subject code
1601
Research subject group en
Industrial engineering
Research subject group fr
Génie industriel
Research subject en
Operations research and management science
Research subject fr
Recherche opérationnelle et science de la gestion
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Combined Statistical Learning and Robust Optimization
Keywords
bayesian statistics; dynamic programming; robust optimization; maintenance engineering; inventory control
Application summary
Most optimization models in industrial engineering account for data with a statistical model, and decisions are made under the assumption the model is correct. On the other hand, models that account for possible model miss-specification, often formulated as robust optimization models, do not account for learning. Such models essentially assume that data is of no value when it comes to learning the model.
The proposed research program aims to develop new models in which a decision maker can account for possible model miss-specification while also learning at the same time. Three streams of research will be investigated.
1. Robust Maintenance Optimization with Learning. Maintenance optimization with learning and model miss-specification has not been considered in the literature. This stream will develop new models that account for learning from data and generate policies that work well even when the underlying model is flawed. The theory will be applied on real-world sensor data to investigate the performance of such models, and to develop visualization and recommendation tools.
2. Robust Inventory Control with Learning. Multi-stage inventory control models with learning and model miss-specification will be developed. While preliminary result shows that the myopic-optimal order quantity is "boosted" when accounting for learning and reduced when accounting for model miss-specification, more research is needed to better characterize these adjustments, and to conduct rigorous validation tests using datasets from Canadian partner companies.
3. Optimization with Multiple Biased Data. A class of multi-source Bayesian control problems wherein the decision-maker can acquire information from multiple, possibly biased sources before making a decision will be investigated. The model will find value in healthcare settings where decisions are based on combining data from different, possibly biased sources. For example, in medical diagnosis practice, it is common to use multiple screening tests or tools (e.g., X-ray, blood work, CT scan) to increase test accuracy and reliability.
If successful, the proposed research will not only provide practical value to companies but lead to new insights explaining the underlying mechanism by which one trades off between learning and model miss-specification. We will disseminate our findings through publications, conferences and practitioner--oriented industry councils. Doctoral students will be trained as future scholars, while those entering industry will champion data analytics in the private sector or the government. (HQPs account for a majority of the budget.) The broader impact will be the improvement of industry practice and educational programs. The research will be conducted in collaboration with Canadian companies connected with the UBC Sauder, bringing together students, faculty and companies to allow for testing of new ideas on datasets, as well as the direct dissemination of knowledge into classrooms.