GEORGE SHANTHIKUMAR
Richard E. Dauch Chair in Manufacturing and Operations Management, Purdue University
Data Integrated Stochastics: Models and Methods
This tutorial will review the current data integrated approaches for predictive and prescriptive analysis of stochastic systems. In particularly we will review: 1) approaches such as Multi-Armed Bandit, Regularization in Sample Average Approximation and Data Driven Robust Optimization for generating prescriptive solutions to stochastic systems, and 2) some of the Machine Learning approaches used for predictive analysis of stochastic systems. We will then provide a framework for data integrated methodology for predictive and prescriptive analytics for stochastic systems. Specific attention will be paid to overcoming structural and statistical errors. This is achieved through Operational Statistics and Objective Operational Learning which are built on the basis of data integration and cross validation. We will illustrate how, 1) regularization in sample approximation approaches and data driven robust optimization with cross validation relates to Operational Statistics, and 2) multi-armed bandit and machine learning approaches compares to Objectives Operational Learning. Applications in pricing and revenue management, inventory control, queueing systems performance evaluation and staffing in service systems will be demonstrated.
This tutorial will review the current data integrated approaches for predictive and prescriptive analysis of stochastic systems. In particularly we will review: 1) approaches such as Multi-Armed Bandit, Regularization in Sample Average Approximation and Data Driven Robust Optimization for generating prescriptive solutions to stochastic systems, and 2) some of the Machine Learning approaches used for predictive analysis of stochastic systems. We will then provide a framework for data integrated methodology for predictive and prescriptive analytics for stochastic systems. Specific attention will be paid to overcoming structural and statistical errors. This is achieved through Operational Statistics and Objective Operational Learning which are built on the basis of data integration and cross validation. We will illustrate how, 1) regularization in sample approximation approaches and data driven robust optimization with cross validation relates to Operational Statistics, and 2) multi-armed bandit and machine learning approaches compares to Objectives Operational Learning. Applications in pricing and revenue management, inventory control, queueing systems performance evaluation and staffing in service systems will be demonstrated.