Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TC72

5 - Deep Autoencoder with Regularization on Sensor Signals for Virtual Metrology in Semiconductor Manufacturing Jeongsub Choi, Rutgers University, 96 Frelinghuysen Road, Core Building, Room 201, Piscataway, NJ, 08854, United States, Harshit Bokadia, Myong K. Jeong In semiconductor manufacturing, feature extraction from raw sensor signals on process equipment is an essential task to build an accurate predictive model for virtual metrology. Autoencoder is a neural network based feature extraction model that compress inputs into features in a latent space. In this talk, we present a new virtual metrology method with a new regularized deep autoencoder considering the characteristics of sensory data. Experimental results with real-life data show that the proposed method improves the performance of prediction models for virtual metrology. n TC70 West Bldg 106B Highly Reliable Complex System Modelling and Analysis Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Dan Yu Co- Chair: Qingpei Hu Co- Chair: Yimin Zhou 1 - Accelerated Degradation Testing Modelling and Analysis with Recovery Capability Considered Qingpei Hu, Chinese Academy of Sciences, 55 East Zhonguancun, Haidian, Beijing, 100190, China, Chengjie Wang, Dan Yu, Jian Liu The precision of ADT models heavily relies on the consistency of the testing environments and the corresponding assumptions. Recovery phenomenon is always observed when measurement conducted offline at low stress for some typical degradation failure modes. Statistical model and inference are proposed for this phenomenon, together with simulation results addressed. 2 - WCF-Approach for Complex System Reliability Assessment Dan Yu, Chinese Academy of Sciences, Beijing, China System reliability assessment from component failure data is usually approximated by the central limit theory combined with delta method. A novel polynomial adjustment method for high reliability systems is adopted to construct higher-order approximate lower confidence limits, elaborated for log-location- scale family models with numerical studies conducted. n TC71 West Bldg 106C Geospatial Optimization Problems Sponsored: Computing Sponsored Session Chair: John Gunnar Carlsson, University of Southern California, Los Angeles, CA 90089, United States 1 - Equitable Partitioning of a Delivery Service Region with Orthogonal Blocks Mehdi Behroozi, Boston, MA, United States Motivated by the last-mile delivery service operation, we consider the problem of utilizing a fleet of vehicles over a rectilinear polygonal service region with orthogonal blocks and fixed vehicle depot locations so that all demand points are serviced and the workload is evenly distributed among the vehicles. We present an efficient algorithm for partitioning the territory into sub-regions using horizontal and vertical lines such that each sub-region contains exactly one vehicle depot location and the workloads of the vehicles, i.e. total traveled distance of the vehicle with respect to some natural metric, are balanced. 2 - The Covering Path Problem on a Grid Liwei Zeng, Northwestern University, Evanston, IL, United States, Karen Smilowitz, Sunil Chopra This paper introduces the covering path problem on a grid (CPPG) which finds the cost-minimizing path connecting a subset of points in a grid such that each point is within a pre-determined distance of a point from the chosen subset. We leverage the geometric properties of the grid graph which captures the road network structure in many transportation problems. We develop a trade-off constraint which quantifies the trade-off between path length and stop count and provides a lower bound for the bi-objective optimization problem. We introduce simple construction techniques to provide feasible paths that match the lower bound within a constant factor.

3 - Distributionally Robust Travelling Salesman Problems and Variations

John Gunnar Carlsson, University of Southern California, 3750 McClintock Avenue, Los Angeles, CA, 90089, United States Recent research on the robust and stochastic travelling salesman problem and the vehicle routing problem has seen many different approaches for describing the region of ambiguity, such as taking convex combinations of observed demand vectors or imposing constraints on the moments of the spatial demand distribution. One approach that has been used outside the transportation sector is the use of statistical metrics that describe a distance function between two probability distributions. Motivated by a districting problem in multi-vehicle routing, we consider a distributionally robust version of the Euclidean travelling salesman problem in which we compute the worst-case spatial distribution of demand against all distributions whose Wasserstein distance to an observed demand distribution is bounded from above. 4 - Fair Division Approaches to Political Districting Gerdus Benade, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, Ariel Procaccia Gerrymandering and the problem of dividing a state into political districts are attracting a lot of attention. One of the difficulties in detecting gerrymandering is that it is hard to determine how many districts a party may expect to win. We approach districting as a fair division problem and propose a target number of districts for every party that is sensitive to the distribution of voters and the geography of the problem. We determine the extent to which this target can simultaneously be guaranteed to every party and propose a computational approach to districting. 5 - Real World VRP Needs a GIS Heather Moe, Product Engineer, Esri, 380 New York St, Redlands, CA, 92373, United States, Shubhada Kshirsagar The Vehicle Routing Problem (VRP) becomes much more complex when the solutions need to be deployed in the real world. To produce an operational result, the VRP model must include geographically precise inputs. The best way to integrate this massive amount of information is to use a Geographic Information System (GIS) in the problem-solving methodology. This presentation will discuss requirements from a wide range of commercial and public-sector clients over 30 years?. With the described GIS inputs, the VRP model will produce a route plan that is both cost efficient and realistic. n TC72 West Bldg 211A Economics of Cybersecurity General Session Chair: Juhee Kwon, Korea Advanced Institute of Science & Technology (KAIST), Seoul, Korea, Republic ofr 1 - Benefit Ambiguity and Asymmetric Herding in Privacy Decision: A Mobile Field Experiment Youngsok Bang, PhD, The Chinese University of Hong Kong, Hong Kong, Jaehyeon Ju, Dong-Joo Lee, Jae-Hyeon Ahn We conduct a field experiment with a self-developed mobile app to examine how the benefit ambiguity of using the app and others’ permission setting affect users’ permission on each of sixteen information items requested by the app. We find that participants who received the detailed reward scheme permitted more items than participants who received the abstract scheme. We also find that participants who permitted fewer items than the average were encouraged to permit more by getting the feedback of others’ permission status, whereas participants who permitted more than the average were not affected. The asymmetric herding effect was greater for participants who received the detailed reward scheme. 2 - Does Sharing Make My Data More Secure? An Empirical Study on Health Information Exchanges and Data Breaches Leting Zhang, Temple University, Philadelphia, PA, 19122, United States, Sunil Wattal This paper examines the IT security implications of hospitals joining a Health Information Exchange (HIE). Using IT governance theory and institutional theory, we propose hypotheses regarding the relation between a hospital joining an HIE and the likelihood that it suffers data breaches. We analyze a panel data for hospitals over a 6 year period empirically. Our preliminary results show joining HIEs mitigates security breach risks for hospitals. We also find that the mitigation effect is higher for hospitals with higher security capabilities and lower for large hospitals. The study contributes to the literature in IT governance, information security, as well as provides practical implications.

333