Informs Annual Meeting 2017

WA13

INFORMS Houston – 2017

4 - Independent Contractors or Full-time Employees? On Staffing a Blended Workforce. Rouba Ibrahim, University College London, MS& I Department, UCL, Gower Street, London, WC1E 6BT, United Kingdom, rouba.ibrahim@ucl.ac.uk, Jing Dong The rise of the blended workforce is identified as one of the top workplace trends in 2017. A blended workforce melds independent contractors or freelancers with permanent employees. We study optimal sta ng decisions in service systems with a blended workforce, and characterize the pertinent trade-offs. 332B Simulation and Optimization Contributed Session Chair: Keli Feng, South Carolina State University, Orangeburg, SC, United States, kfeng1@scsu.edu 1 - A State Dependent Inventory Policy for Multi Period, Multi Stage Systems Anna-Lena Sachs, Professor, University of Cologne, Albertus-Magnus-Platz, Cologne, 50923, Germany, anna-lena.sachs@uni-koeln.de, Stewart Liu, Stefan Minner, Philip Kaminsky Inventory management for perishable products such as fruit and vegetables is a challenging task for retail managers. We investigate a retail chain that owns several stores and keeps inventory both at the stores and at a central warehouse. We determine the optimal inventory policy as a function of external variables such as price, weekday and weather. By integrating forecasting and inventory optimization using robust optimization and a gradient-search algorithm, we are able to solve this complex problem for a large number of stores and variables 2 - Multiperiod Inventory Management with Advance Supply Contracts Sercan Demir, Teaching/Research Assistant, University of Miami, 1251 Memorial Drive #268, Department of Industrial Engineering, Coral Gables, FL, 33146, United States, s.demir@umiami.edu, Mazhar Arikan, Murat Erkoc We study a multi-period inventory control problem of a seasonal product where the demand for the product is stochastic and non-stationary across periods. The demand distribution across the planning horizon is shaped by a market signal that is random at the beginning of the planning horizon. The firm has the option of contracting its supplies before the realization of the market signal to benefit from price discounts. It is also possible for the firm to replenish its inventories during the selling season by paying higher spot market prices. Our research is motivated by applications in cruise line industry where contracts are made before the final number of bookings is realized. 3 - A Two-phase Derivative-free Optimization Algorithm for a Process Governed by Non-linear Algebraic Partial Differential Equations Ishan Bajaj, Texas A&M.University, 503 Cherry St., College Station, TX, 77840, United States, ishan.bajaj@tamu.edu, Shachit S. Iyer, Priyadarshini Balasubramanian, Faruque Hasan A data-driven two-phase trust-region based optimization algorithm is proposed to solve problems where the constraints and objective function cannot be expressed analytically in terms of decision variables. In the first phase of the algorithm, a feasible point is obtained and in the second phase, the objective function is decreased while maintaining feasibility. The algorithm is applied to optimize the design and operating conditions of an integrated carbon capture and conversion process such that constraints related to product quality and emissions are satisfied and cost is minimized. 4 - Continuous Inspections Assessing Economic Feasibility of WA12

5 - Bayesian Optimization with Gradients Matthias Poloczek, Cornell University, 272 Rhodes Hall, 136 Hoy Road, Ithaca, NY, 14853, United States, poloczek@cornell.edu, Jian Wu, Andrew Wilson, Peter Frazier Bayesian optimization has proven successful for global optimization of expensive- to-evaluate, multimodal functions. However, it typically does not use derivative information. In this talk, we show how Bayesian optimization can exploit derivative information to decrease the number of objective function evaluations required for good performance. In particular, we develop a novel algorithm, the derivative-enabled knowledge-gradient, that provides state-of-the-art performance compared to a wide range of optimization procedures with and without gradients, on benchmarks including logistic regression, kernel learning, and k-nearest neighbors. 6 - Performance Comparison under Two Replenishment Policies in Master Production Scheduling with Demand Uncertainty Keli Feng, Associate Professor, South Carolina State University, 300 College Street NE, Department of Business Administration, Orangeburg, SC, 29117, United States, kfeng1@scsu.edu We study a single end-product master production scheduling problem with supply capacity limits and dynamic stochastic demand. Our replenishment options include using an order quantity vector or using order-up-to levels. The performance measures we consider include service levels, average cost, standard deviation of cost, expected upside risk, “cost-at-risk”, and probability of exceeding a target cost. Using simulation-based optimization method, we perform a computational experimental design to explore the impact of parameters on performance measures under both replenishment policies. 332C Decision Support Systems Contributed Session Chair: Gizem Erdinç, erciyes university, kayseri, Turkey, gizemerdincc@gmail.com 1 - Using a Moving Reference Approach for Measuring Balance in Resource Allocation Problems Hale Erkan, Bilkent University, Bilkent University, Engineering Department, Ankara, 06800, Turkey, hale.erkan@bilkent.edu.tr, Ozlem Karsu We consider a resource allocation settings where there are several categories of projects, each with uncertain cost and benefit values. The decision maker (DM) tries to determine the projects to fund given a limited budget. The two objectives are maximizing efficiency and balance among categories. Balance is measured with respect to a reference distribution (RD), which changes as used budget changes. We propose a biobjective robust optimization model and find Pareto solutions using the E-constraint method. Our results show that considering balance explicitly provides useful insights to the DM. 2 - Context-aware Process and Organization Analytics: Extending Business Analytics Towards More Effective and Flexible Organization Seunghoon Lee, PhD Candidate, POSTECH, 77, Cheongam-ro Dept. of Industrial and Management Engineering, Pohang, 37673, Korea, Republic of, frhyme@postech.ac.kr, Injun Choi Recently, as related technologies such as machine learning has been advanced rapidly, interest in business analytics (BA) has been significantly increased in academia and industry. However, existing BA approaches only consider organization from limited perspective thus fail to improve the actual performance of organization from holistic view point. Thus, we propose a new BA paradigm by developing the Context-aware Process and Organization Analytics to flexibly cope with various contextual changes in the complex environment. In particular, we define an ontology that can represent situational knowledge of the organization by identifying the relationships among organizational elements. WA13 The construction industry serves humanity in a subjective and risky environment in comparison with other industries. In contrast to automation, almost every processes are done by human. Therefore, project managers should overcome much more uncertainties and probability in solving management problems of construction projects. A risk analysis was presented while adopting some criteria were affect the outcome performance of project such as cost, time quality, performance and safety. The impact levels of criteria and emergence ratings were calculated using Fuzzy Sets and ranked. In this way, decision makers who work in management level in companies can make better and more accurate decisions. 3 - Fuzzy Risk Analysis for Construction Projects Gizem Erdinç, Erciyes University, Kayseri, Turkey, gizemerdincc@gmail.com, Feyza Gurbuz

Strategies under a Changing Demand Distribution Anh Ta, University of North Texas, Denton, TX, 76201, United States, Anh.Ta@unt.edu, Robert Pavur

Continuous inspections is an essential part of many processes. Minimizing the long run cost is an important criterion in selecting a strategy. This research assesses proposed continuous auditing methodologies by using a Monte Carlo Simulation approach under conditions that may be changing over time. Estimating the parameters for the rate of occurrences is an important step in using effective continuous inspection techniques. Applications to using this methodology in supply chain inspections are examined.

445