2016 INFORMS Annual Meeting Program

TA38

INFORMS Nashville – 2016

TA38 206A-MCC Reliability II Contributed Session Chair: Mengmeng Zhu, Rutgers University, Piscataway, NJ, 8854, United States, mengmeng.zhu@rutgers.edu 1 - Optimal Design Of Hybrid Sequential Testing For A System With Mixture Of One-shot Units Elsayed A Elsayed, Distinguished Professor, Rutgers University, 96 Frelinghuysen Rd, Piscataway, NJ, 08854-8018, United States, elsayed@rci.rutgers.edu Non Destructive Testing is conducted to determine the functionality of the units without permanent damage in order to estimate the units’ reliability. In this presentation, we investigate a system composed of non-identical units with different characteristics and subjected to hybrid reliability testing (Destructive and NDT). It is of interest to optimally design the hybrid sequential reliability testing. After conducting a number of hybrid testing, we decrease the sample size of the destructive testing as the accuracy of reliability metrics estimation improves. Eventually, we only need to conduct NDT only. The efficiency and accuracy of the proposed methods are validated. 2 - Transportation Network Fragility And Economic Losses Narges Kaveshgar, University of South Carolina, 300 Main Street, Department of Civil and Environmental Engineering, Columbia, SC, 29208, United States, kaveshga@email.sc.edu, Nathan Huynh, Joseph Von Nessen Interdependencies between the transportation system and other critical infrastructures necessitate the need to protect it to achieve system resiliency. Current study develops a methodology to quantify the robustness and investigate the reliability of transport network under extreme events. To this end, perishable field data is collected to determine the impact of the recent road and bridge closures caused by historic rainfall event in South Carolina to the traveling public and businesses. 3 - Analyzing Coastal Highway Network Reliability In Hurricane Flooding Surge Through Geographic Information System Lei Bu, Institute for Multimodal Transportation, Jackson, MS, United States, leibu04168@gmail.com, Feng Wang Reliability is related to the ability of a network to carry out desire traffic flow which includes node to node blocking or delay. Based on the history data of hurricane flooding surge in Gulf coastal region, coastal highway network reliability is analyzed using a geographic information system. Spatial statistics and analyst methods based on density, geographic distribution and bilinear interpolation are used to calculate point density, Z score and hot spots for the hurricane flooding surge data. Based on the spatial statistics and analyst results, the blocking or delay links, namely, nodes to nodes, on highway network are found to determine the network reliability. 4 - Two Phase Imperfect Debugging And Imperfect Fault Removal Software Reliability Modeling Mengmeng Zhu, Rutgers University, Piscataway, NJ, 08854, United States, mengmeng.zhu@rutgers.edu, Hoang Pham A software reliability modeling considering software fault type and multi-phase debugging process is proposed in this paper. Type I fault and Type II fault represent independent and dependent software fault during debugging, respectively. Two-Phase debugging process are discussed in the model development. Additionally, a small portion of software faults that software testers are not able to remove is included in this study due to the limitation of resource and knowledge. TA39 207A-MCC Artificial Intelligence in Big Data General Session Chair: Xiao Liu, University of Arizona, 1300 E. Fort Lowell Road G109, Tucson, AZ, 85719, United States, xiaoliu@email.arizona.edu 1 - Does Interim Winner’s Performance Information Playa Role? An Empirical Investigation Of The Rank-order Newsvendor Contests Abraham Seidmann, Simon Business School, University of Rochester, Simon Business School, Dir of OR Dept, Rochester, NY, 14627, United States, avi.seidmann@simon.rochester.edu, Tong Wu Many firms award bonuses to their employees based on their relative performance. When facing newsvendor-type decisions under this type of inter- worker competition, firms need to consider what type of information to disclose to the employees from period to period in order to achieve better outcomes in the long run. Using a laboratory experiment, we find that publicly displaying the

winner’s performance information every period can significantly improve individuals’ overall newsvendor decision making compared to the control group, although the pull-to-center effect is observed. Using another experiment, we find

that impulsivity can explain the observed pull-to-center bias. 2 - Big Data In The Healthcare And Wellness Industry

Stephen J Stoyan, Director, Business Analytics and Strategy, Abbott Laboratories, 100 Abbort Park Road, Chicago, IL, 60064, United States, stephen.stoyan@abbott.com Today’s healthcare and wellness industry is dynamic, competitive, and global demands require extremely high volume. Keeping your supply chain lean and efficient is imperative to driving cost savings. Staying competitive requires a sales campaign that is connected to customers at new levels. Big data and advanced analytics are integral parts of the business that provide supply chain efficiencies and top line growth through strategic and operational initiatives. Analytically tuned tools are discovering new opportunities, making connections, and creating business value streams in areas not well understood. We present big data initiatives at Abbott Laboratories and their impact on the business. 3 - Sales Assistance Search And Purchase Decisions An Analysis Using Retail Video Data Aditya Jain, Baruch College, Zicklin School of Business, 55 Lexington Ave, Suite 9-240, New York, NY, 10010, United States, aditya.jain@baruch.cuny.edu, Sanjog Misra, Nils Rudi We investigate the roles of sales assistance and search in driving customer’s purchase decision using unique observational video data from retail stores. Our analysis reveals that both sales assistance and search play substantial roles which differ based on the context of specific decisions—search has a more dominant role in purchase incidence, whereas the latter in conditional expenditure. 4 - Mining E-cigarette Adverse Events Using The LSTM-based RNN Model With Word Embeddings Features Jiaheng Xie, University of Arizona, Department of Management Information Systems, Tucson, AZ, 6, United States, xiej@email.arizona.edu The past years have witnessed increased popularity of e-cigarette use across the world. However, the risk of cartridge fluids and emissions is relatively under- examined due to limited user sample size. Social media provide a large corpus that contains e-cigarette related information. In order to study the e-cigarette adverse effects in a more comprehensive manner, we propose to study the adverse events of e-cigarette with a large volume of health social media data. The challenges in e-cigarette safety social media monitoring lie in identifying relevant adverse events reported by consumers in noisy social media content with high accuracy. The current automatic entity recognition methods have unsatisfying performance due to consumer vocabulary used in social media. To address this issue, we developed a Long Short Term Memory (LSTM) based Recurrent Neural Network (RNN) model to extract the medical entities. Based on our results, our proposed LSTM-based RNN model with word embeddings achieved better entity extraction performance, with a precision of 90.58%, recall of 82.43% and f-score of 86.31%. We identified 1,212 adverse event entities, 397 e-cigarette component entities (chemicals, flavors, and brands) and the corresponding component-event relationships. Since certain e-liquid chemicals, flavors and e-cigarette brands are significantly associated with adverse events, regulatory actions are in need. Certain flavors and brands should also be controlled due to their adverse events.

TA40 207B-MCC Markov Decision Processes: Applications Sponsored: Applied Probability Sponsored Session

Chair: Jie Ning, Case Western Reserve University, 11119 Bellflower Rd, Case Western Reserve University, Cleveland, OH, 44106, United States, jie.ning@case.edu Co-Chair: Matthew J Sobel, Case Western Reserve University - Retired, 11119 Bellflower Rd, Case Western Reserve University, Cleveland, OH, 44106, United States, matthew.sobel@case.edu 1 - Optimal Policies For Risk-averse Electric Vehicle Charging With Spot Purchases Daniel Jiang, University of Pittsburgh, Pittsburgh, PA, 15261, United States, drjiang@pitt.edu, Warren B Powell We consider the sequential decision problem faced by the manager of an electric vehicle (EV) charging station, who aims to satisfy the charging demand of the customer while minimizing cost. We formulate the problem as a finite horizon Markov decision process (MDP) and provide an analysis of the effect that risk parameters, e.g., the risk-level used in CVaR, have on the structure of the optimal policy. We show that becoming more risk-averse in the dynamic risk measure sense corresponds to the intuitively appealing notion of becoming more risk-averse in the order thresholds of the optimal policy.

246