2016 INFORMS Annual Meeting Program

WC08

INFORMS Nashville – 2016

WC06 102A-MCC Big Data 2 Sponsored: Data Mining Sponsored Session

3 - Observational Data Driven Modeling And Optimization Of Manufacturing Processes Najibe Sadatijafarkalaei, PhD Student, Wayne State University, 4815 Fourth Street, Detroit, MI, 48202, United States, fv0017@wayne.edu, Ratna Babu Chinnam The main objective of this study is to rely on observational data to achieve robust parameter design of manufacturing processes. Controlled experiments can be challenging in production environments and this paves for an effective alternative approach to attain robust process parameter conditions. The proposed framework relies on an integrated feature selection, response surface modeling, and optimization methodology. We also report illustrative results from a tire compound production process.

Chair: Milton Soto-Ferrari, Western Michigan University, 4601 Campus Drive, Parkview Campus Building I, Kalamazoo, MI, 49008-5336, United States, miltonrene.sotoferrari@wmich.edu 1 - Disease Detection Analytics: Comparing And Contrasting The Performance Of Popular Predictive Models For Breast Cancer And Diabetes Data Sets

WC08 103A-MCC

Subhashish Samaddar, Professor, Business Analytics and Operations, Georgia State University, Managerial Science, P.O. Box 4014, Atlanta, GA, 30302-4014, United States, s- samaddar@gsu.edu, Somnath Mukhopadhyay

Technology Mgt Contributed Session

Disease detection based on clinical data of the patient can save health care costs. Consequently, data mining in disease detection is fast gaining popularity. Our study applies popular predictive modeling algorithms such as random forest, linear programming classifiers, and, neural network to two examples data sets of clinical data: Breast cancer and Diabetes. The former data set has been bench marked in the literature - so comparison with prior results was possible. The data set on Diabetes can be used as new benchmark for future research. The article reports compares and contrasts results from each method. 2 - Characterization Of Breast Cancer Patients Receiving Unexpected Treatments Milton Soto-Ferrari, Western Michigan University, 4601 Campus Drive, Parkview Campus Building I, Kalamazoo, MI, 49008-5336, United States, miltonrene.sotoferrari@wmich.edu, Diana Prieto In 2016, approximately 40,450 women in the US are expected to die from breast cancer. Medical treatments are mainly driven by clinical factors including cancer staging, tumor size, histology, and age. This research aims to propose a systematic methodology to identify the clinical and non-clinical features that influence the receipt of an unexpected treatment in breast cancer patients. We extend the factor exploration and characterization of patients through a Bayesian Network breakdown methodology that allows the analysis of conditional probabilities to relate patient features with treatment receipt. We use registers of the SEER program from Detroit area considering the period 2007-2012. Process Modeling Sponsored: Data Mining Sponsored Session Chair: Najibe Sadatijafarkalaei, Wayne State University, 4815 Fourth Street, Detroit, MI, 48202, United States, fv0017@wayne.edu 1 - An Efficient Nonparametric Fault Variable Identification Method Mehmet Turkoz, Rutgers University, 16 Rachel Terrace, Piscataway, NJ, 08854, United States, turkoz@scarletmail.rutgers.edu In a process, identifying which variables cause an out-of-control signal is a challenging issue for quality problems. If the distribution of the process is unknown, existing parametric methods are not suitable for identification of changed variables. In this paper, we propose a new nonparametric method to identify the fault variables and demonstrate its performance through various simulation studies. 2 - A Hybrid Genetic Algorithm With Tabu List For Generating a Stochastic Process Tree Model Based on Event Logs Jin Young Choi, Ajou University, Worldcup-ro 206, Yeongtong-gu, Suwon, 16499, Korea, Republic of, choijy@ajou.ac.kr, Woo-Min Joo, Do Gyun Kim We present an efficient hybrid algorithm integrating genetic algorithm and tabu search for generating a stochastic process tree model using event logs. It is examined for its performance by considering some example event logs in literature, evaluating four fitness measure such as simplicity, precision, replay, and generalization. WC07 102B-MCC

Chair: Mahmut Sonmez, Senior Lecturer in Management Science & Statistics, University of Texas at San Antonio, College of Business, San Antonio, TX, 78249-0631, United States, maho.sonmez@utsa.edu 1 - Wearable Technology In Fitness – Fitbit Hongwei Du, Professor, California State University-East Bay, 25800 Carlos Bee Boulevard, Hayward, CA, 94542, United States, hongwei.du@csueastbay.edu One trending of wearable technology is Fitness Devices. This paper focuses on wearable technology for fitness tracking and the FitBit Company. It identifies and presents wearable technology behind the FitBit products, the pros and cons of the FitBit, and the role that FitBit plays in the Internet of Things. Last, the future of the FitBit Company and product is discussed. 2 - Dominant Design, Sequential Product Categories, And Product Innovation Hyunwoo Park, Postdoctoral Fellow, Georgia Institute of Technology, 85 5th St NW, Atlanta, GA, 30332, United States, hwpark@gatech.edu, Rahul C Basole We study the impact of dominant design in sequential product categories on product innovation using a dyadic perspective in the context of mobile phone industry. Our results indicate that dominant design accelerates incremental product innovation and causes temporary adverse shift in product category focus. 3 - Technological Innovation, International Patenting And National Economic Development: A Multinational Multi-year Study Kelvin Wayne Willoughby, Professor, Innovation and Intellectual Property, Skolkovo Institute of Science and Technology, Skolkovo Innovation Center, 3 Nobel Street, Moscow, 143026, Russian Federation, kelvin@skoltech.ru, Alexander Vidiborskiy This paper reports the results of a study of the inbound and outbound patenting activity of 78 countries for which reliable data were obtained over the 14 years from 2000 to 2013. Several new indicators of the international patenting proclivities of inventors were utilized in multiple phases of statistical analysis and data vizualization to investigate the relationship between domestic inventive activity, international patenting profiles and changes in the relative per capita wealth levels of countries. The results suggest that the economic benefits countries may gain from investing in technological innovation may be enhanced by emphasizing the international patenting of domestic inventions. 4 - The D-day, V-day, And Bleak Days Of A Disruptive Technology: A New Model For Ex-ante Evaluation Of The Timing Of Technology Disruption Chialin Chen, Professor, National Taiwan University, Taipei, 10617, Taiwan, cchen026@ntu.edu.tw, Jun Zhang, Ruey-Shan Guo We conduct theoretical and empirical analyses to evaluate the timing of technology disruption. We conceptualize the ease and network factors as key determinants of performance improvement for a disruptive technology. A dynamic consumer model is developed to identify two critical times, termed D- Day and V-Day, of technology disruption. We also show that there may exist some “bleak days” during which a firm would discontinue a “promising” technology that will eventually disrupt. Empirical tests are conducted with data of hard disk drives, semiconductor technologies, and CPU performance for mobile devices to verify key model assumptions and to show how to estimate the ease and network factors.

429