2016 INFORMS Annual Meeting Program

TA04

INFORMS Nashville – 2016

TA03 101C-MCC

3 - Productivity Assessment Through Knowledge Generation In Automotive Manufacturing Sector, Utilizing Statistical Methods Amir Abolhassani, West Virginia University, 1233 Pineview Drive, Apt 14, Morgantown, WV, 26505, United States, Amirabolhassani@gmail.com, Bhaskaran Gopalakrishnan The research investigates current strategies that help automobile manufacturers to enhance their productivity. The study utilizes statistical methods to define the most important effective factors on the most well-known productivity measurement, Hours per Vehicle (HPV), in the automotive industry in North American manufacturing plants. 4 - A Machine Learning Approach For Developing Intersection Safety Performance Function Hamidreza Ahady Dolatsara, Auburn University, Suite 3301, 345 West Magnolia Ave., Auburn, AL, 36849, United States, hamid@auburn.edu, Fadel Mounir Megahed This study introduces a non-subjective method for identifying intersection related crashes based on their distance to the center of intersection. Then utilizes data of those crashes for developing Safety Performance Functions (SPFs). This study utilizes machine learning algorithm to investigate histogram of crashes’ distance to the center of intersections. The histogram classified to intersection and non- intersection related parts. After classifying the crashes, intersection SPFs are developed. Chair: Seth Guikema, University of Michigan, 1205 Beal Ave., Ann Arbor, MI, 48109-2117, United States, sguikema@umich.edu 1 - Adverse Event Prediction: Forecasting Wind Power Ramps Andrea Staid, Sandia National Labs, Albuquerque, NM, United States, astaid@sandia.gov Wind power ramp events (large changes in output over a short period of time) are of particular concern in power systems with high wind penetration. They are also often difficult to predict. We present statistical methods for combining multi- source data to better predict the adverse ramp events that are typically not captured in a standard weather forecast. We present a case study using data from the Bonneville Power Administration and focus on farm-specific ramps. 2 - Predicting Low-wind Events To Inform Planning And Policy Incentives Kristen R. Schell, University of Michigan, Ann Arbor, MI, United States, krschell@umich.edu, Seth Guikema Wind power is currently the most cost-effective source of renewable energy, from the perspective of life-cycle cost estimates. Given major government backing to support the general expansion of renewable energy, wind power continues to be the investment of choice for developers, largely due to its generally higher resource potential and capacity factors. Hence, as wind power becomes increasingly integrated into the electric grid, the power system in turn becomes increasingly vulnerable to “wind droughts”, or periods of low-to-no wind power. This study uses large-scale wind data to predict low-wind events, to better inform system planning and renewable policy incentives. 3 - Learning Of Imbalanced Data For Predicting The Power Outage Elnaz Kabir, PhD Student, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109-2117, United States, ekabir@umich.edu, Seth Guikema In this article we want to model the number of power outages during the hurricanes. Making accurate prediction of power outage can be really valuable for the utility companies as well as customers and public agencies to make better response planning. Our data-set is highly zero inflated and imbalanced because power outages occur rarely. Since no techniques for dealing with imbalanced data sets are consistently better for all conditions, we investigate several methods to find the most appropriate strategy for our data set. TA02 101B-MCC Predictive Modeling for Wind Power Sponsored: Data Mining Sponsored Session

Entertainment Analytics Invited: Entertainment Analytics Invited Session Chair: Christian Peukert, University of Zurich, Rämistrasse 71, Zurich, 8006, Switzerland, christian.peukert@uzh.ch 1 - Freemium Pricing: A Stylized Framework And Evidence From A Large-scale Field Experiment Jörg Claussen, LMU Munich, Munich, Germany, j.claussen@lmu.de, Jörg Claussen, Copenhagen Business School, Frederiksberg, Denmark, j.claussen@lmu.de, Julian Runge, Julian Runge, Stefan Wagner Marketers struggle to find optimal designs for their freemium offerings. We present a stylized framework of freemium pricing that systematizes key choice variables and identifies their interaction with relevant outcome variables. Firms set the share of free product features and the price of premium content and implement viral mechanisms. These choices affect monetization not only via their effect on users’ conversion to paying customers, but also via their effects on usage behavior and viral activities. We apply our framework to a large-scale field experiment and show that a reduction of free product features increases conversion rates and viral activities, and has ambivalent effects on usage. 2 - Building An Online Reputation With Free Content: Evidence From The E-book Market Dainis Zegners, LMU Munich, d.zegners@lmu.de An important strategy to build a reputation is to offer products for free to induce buyers to provide feedback. I argue that giving away free products to build a reputation can be a double-edged strategy. It does not only attract buyers with a high preference, but also buyers with a low preference, who give worse feedback. I test the strength of this negative effect on reputation using data from an online platform where I observe self-published authors either selling their e-books at a positive price or giving them away as free content. Consistent with a negative selection effect on reputation, I observe that buyers who receive an e-book as free content rate it worse than buyers who buy it at a positive price. TA04 101D-MCC Advances in Using Market Models for Power Transmission Planning Sponsored: Energy, Natural Res & the Environment, Energy I Electricity Sponsored Session Chair: David Pozo, Pontificia Universidad Catolica, Avendia Vicuna Mackenna #4860, Puebla del Principe, 7820436, Chile, davidpozocamara@gmail.com 1 - Non-cooperative Multi-regional Transmission Planning Saamrat Kasina, Johns Hopkins University, Ames 214, Baltimore, MD, 21218, United States, bkasina1@jhu.edu, Benjamin Field Hobbs Traditional transmission planning methods overlook the political boundaries within which planning entities operate. However, in reality, there are multiple regional transmission planning agencies with limited coordination with each other during planning. We develop a bi-level model that defines the relationship between multiple non-cooperating transmission planners, generation investors, and the energy market equilibrium. We ask how the transmission plans from such a process differ from those from a cooperative planning process and what the value of cooperation is, if any. 2 - Bi-level Network Planning Model Considering Generation-market Equilibria Subject To Inefficient Network Pricing We consider proactive planning of transmission subject to the response of a generation market with imperfect long-run transmission pricing based on a MW- km charging system, similar to that of the UK and elsewhere. MW-km-based charging distorts siting decisions of both thermal and renewable plants relative to the social optimum because (1) congestion and locations of needed reinforcements are ignored and (2) all generation types at a location pay the same charge, no matter how they use the grid. Pengcheng Ding, Johns Hopkins University, Baltimore, MD, United States, pangchingting@gmail.com, Benjamin Hobbs

233