2015 Informs Annual Meeting

WC31

INFORMS Philadelphia – 2015

3 - The Impact of Digital Natives on B2B Purchasing Decisions Haris Krijestorac, PhD Student, Information Systems, University of Texas at Austin, 3106 Speedway, Apt. B203, Austin, Tx, 78705, United States of America, haris.krijestorac@utexas.edu Our study seeks to better understand how businesses buys from businesses in a marketplace with increasingly digitally oriented decision makers. We analyze how ‘Digital Natives’ and ‘Digital Immigrants’ differ when influencing purchasing decisions in a business-to-business (B2B) context. To this end, we are surveying purchase decision makers at B2B companies. The results of our survey can help B2B companies better understand their target market, and refine their approach to marketing and sales. 4 - Improving The Persuasive Effectiveness of Anti-piracy Educational Campaign Messages Bong Keun Jeong, Assistant Professor, Metropolitan State University of Denver, Campus Box 45, P.O. Box 173362, Denver, CO, 80217, United States of America, bjeong@msudenver.edu, Tom Yoon Prior literature suggests that anti-piracy educational campaign is an effective way to dissuade users from downloading illegal contents. However, most comments and opinions on public campaigns are against what they seek to achieve. The objective of this study is to explore ways to improve the effectiveness of anti- piracy campaigns. We examine the impact of message frame, issue involvement, risk perception, and message evidence on the persuasive effectiveness of anti- piracy campaign messages. 5 - Collective Innovation in 3D Printing: Novelty, Reuse and Their Interplay Harris Kyriakou, Stevens Institute of Technology, 1 Castle Point on Hudson, School of Business 642, Hoboken, NJ, 07030, United States of America, ckyriako@stevens.edu, Jeffrey V. Nickerson We present an empirical fitness landscape of open hardware designs and examine how (i) digital artifact reuse, (ii) novelty and (iii) their interaction affect the success of open innovation endeavors. We use mixed research methods to draw insights about how members of open innovation communities build upon preexisting work and collaborate to create something new. A theory of search in the design space is used to explain how these seemingly contradicting forces affect innovation. Chair: Adrian Albert, Senior Data Scientist, C3 Energy, 1300 Seaport Blvd, Suite 500, Redwood City, CA, 94063, United States of America, adrian.albert@c3energy.com 1 - Data-Driven Structural Health Monitoring of Large-Scale Power Grids Adrian Albert, Senior Data Scientist, C3 Energy, 1 Physical asset failure on power grids leads to costly loss-of-service. In the past, energy utilities have only performed maintenance retroactively. Yet new sensing and data processing allow for a near real-time picture of network state. However in practice the sensors used to detect faults often malfunction. Here we describe a system for predicting faults on certain asset types. We also develop a model for optimizing sensor replacement as to ensure desired operation profiles. 2 - Simulating Annual Variation in Load, Wind, and Solar by Representative Hour Selection John Bistline, Technical Lead And Project Manager, Electric Power Research Institute, 3420 Hillview Avenue, Palo Alto, CA, 94305, United States of America, Jbistline@epri.com The spatial and temporal variability of renewables have important economic implications for investment and system operation. This talk describes a method for selecting representative hours to preserve key distributional requirements for regional load, wind, and solar time series with a two orders of magnitude reduction in dimensionality. We discuss the implementation of this procedure in the US-REGEN model and compare impacts on energy system decisions with more common approaches. 3 - C3 Cyberphysix: An Operating System for the Smart Grid Mehdi Massoumy, Senior Data Scientist, C3 Energy, 1300 Seaport Blvd, Suite 500, Redwood City, CA, 94063, United States of America, mehdi.maasoumy@c3energy.com We propose an operating system that aggregates data from disparate data sources across the smart grid value chain, applies analytics on the data, and makes 300 Seaport Blvd, Suite 500, Redwood City, CA, 94063, United States of America, adrian.albert@c3energy.com WC31 31-Room 408, Marriott Data-driven Operations Management of Energy Systems Sponsor: Data Mining Sponsored Session

optimal decisions for the operation of the system. The proposed operating system in real-time analyzes the requirements of the supply side, and requirements of the demand and performs optimal operation of the system while taking into account the constraints of all the components of the grid. 4 - Data-Driven Management of Large, Distributed Energy Systems – The Case of Residential Solar Networks Amir Kavousian, Data Scientist, Sunrun Inc., 747 Anderson St, San Francisco, CA, 94110, United States of America, amirk@alumni.stanford.edu This talk presents the data-driven operation of one of the largest residential solar fleets in the US. I demonstrate how advanced statistical methods are deployed to proactively identify operational issues and their root causes. The data-driven insights are fed back into operations, customer relations management, sales, marketing, product, and design teams. In particular, I explain a novel method to estimate the long-term, gradual decrease in solar systems productivity, known as degradation.

WC32 32-Room 409, Marriott Data Mining Methodology Contributed Session

Chair: Xuelei Ni, Associate Professor, Kennesaw State University, 365 Cobb Ave, Suite 209, MD# 1601, Kennesaw, GA, 30144, United States of America, xni2@kennesaw.edu 1 - Multivariate Statistical Analysis in NPD: Customization of a Sustainable Product Istefani Paula, Professor, UFRGS, Osvaldo Aranha street 99, Porto Alegre, 90035190, Brazil, istefani@producao.ufrgs.br, Manoel Silveira, Angela Marx, Ana Facchini, Márcia Elisa Echeveste The aim of this poster is to present an approach to identify clusters of consumers using Chi squared Automatic Interaction Detector. Based on the clusters formed it is possible to find different segments and to associate requirements demanded by them what allows the customization by means of product derivation. The method contributes to the Requirements Management area illustrated in the development of an eco friendly household cleaning product. 2 - Change Detection using Local Amplitude and Phase Synchronization in Complex Dynamical Systems Ashif Sikandar Iquebal, Texas A and M University, 4501 College Main St, Apt. 1002, Bryan, TX, 77801, United States of America, ashif_22@tamu.edu, Satish Bukkapatnam We propose a novel technique to detect changes in dynamical systems using local phase and amplitude synchronization among its constituent signals, generated using a non-parametric time scale decomposition method. We identify a set of components that is likely to capture the information about dynamical changes of interest using a maximum mutual agreement concept. Finally, a statistic is defined that can be employed to detect changes in complex systems where other methods fail. 3 - Between-Participants’ Discourse Bias in Comments Classification: Adjusting Tf-idf Inbal Yahav, Lecturer, Bar Ilan Business School, Bar Ilan University, Ramat Gan, Is, 52900, Israel, inbal.yahav@biu.ac.il, David Schwartz Text mining has gained great momentum in recent years. A leading research branch in this regard is the field of comment classification. An essential pre-step in comment classification is words processing, commonly achieved by using the tf-idf formula. This work reveals, analyses and correct the bias introduced by between-participants’ discourse to tf-idf. We show that ignoring this bias can manifest in a non-robust method at best, and can lead to an entirely wrong conclusion at worst. 4 - High-Dimensional Semi-Supervised Learning via a Fusion-Refinement Procedure Xuelei Ni, Associate Professor, Kennesaw State University, 365 Cobb Ave, Suite 209, MD# 1601, Kennesaw, GA, 30144, United States of America, xni2@kennesaw.edu, Xiaoming Huo, Zhikun Lei, Renfu Li This paper develops a sufficient dimension reduction (SDR) approach for the high-dimensional semi-supervised learning (SSL) problem. We first modify the fusion-refinement procedure, an SDR technique, to extract the essential features for a lower-dimensional representation, then apply an SSL algorithm in the lower-dimensional feature space to tackle the SSL problem. Numerical experiments demonstrated the effectiveness of the new technique.

436