2015 Informs Annual Meeting

WA06

INFORMS Philadelphia – 2015

WA05 05-Room 305, Marriott Identifying Sentiment Change and Geographic Location in Social Media Cluster: Social Media Analytics Invited Session Chair: Chris Smith, TRAC-MTRY, 28 Lupin Lane, Carmel Valley, 93924, United States of America, cmsmith1@nps.edu 1 - Identifying Changes in Twitter Sentiment Sam Buttrey, Assoc. Prof., Naval Postgraduate School, Code OR/Sb, Monterey, CA, 93943, United States of America, buttrey@nps.edu, Jon Alt This ongoing research demonstrates the application of statistics and machine learning to identify spatio-temporal changes in population sentiment using 10 TB of recent Twitter data. It also seeks to compare Twitter sentiment to results of surveys taken in the same places and times. This comparison may inform uses of sentiment analysis as an alternative to the use of structured surveys in areas where surveying is infeasible. Practical difficulties in analyzing big data of this sort are discussed. 2 - Changes in Network Topography to Predict Social Unrest using Social Media Rob Schroeder, Naval Postgraduate School, 526 Union St., Monterey, CA, 93940, United States of America, rcschroe@nps.edu In recent years, social media has become a common communication medium for social movements. These social movements are able to interact with members, sympathizers, and the general public using social media. This research analyzes how the overall structure of their interactions via Twitter change over time and compares the changes to planned events by the social movement. 3 - Better Defining Location and Attribute Data in Twitter by Utilizing Wikipedia Localization Text Patrick Dudas, Contractor, NPS, 1215 Wisconsin Ave, Pittsburgh, PA, 15216, United States of America, dudaspm@gmail.com Within Twitter understanding users’ geolocation is subject to either the user geotagging their tweets or a high-level profile location. Working with Wikipedia and Twitter, we better define locations and their localized names and attributes by means of Wikipedia’s rich datasets. Parsing Wikipedia, we can produce location objects and their localized translation of the location around the world, producing a better means of understandings both the user’s location and voice on Twitter. 4 - Non-linear Dynamics of Human Emotions: Analysis of Twitter Data and its Implications Les Servi, The MITRE Corporation, 202 Burlington Road, Bedford, MA, United States of America, lservi@mitre.org, Waldemar Karwowski, Dylan Schmorrow, Nabin Sapkota Exploration of the extent that human emotions, expressed in Twitter data, have chaotic and non-linear dynamics has profound implications in its use for forecasting a population’s mood. This study examines such dynamics through the analysis of hundreds of thousands of Twitter messages. WA06 06-Room 306, Marriott Modeling and Computations in Financial Engineering Sponsor: Financial Services Sponsored Session Chair: Lingfei Li, Assistant Professor, The Chinese University of Hong Kong, 608 William M.W.Mong Engineering BLD, Shatin, Hong Kong - PRC, lfli@se.cuhk.edu.hk 1 - Long Term Risk: A Martingale Approach Likuan Qin, Northwestern University, 2145 Sheridan Rd, Tech We extend long-term factorization of the pricing kernel to general semimartingale environments, without assuming the Markov property. We explicitly construct long-term factorization in HJM models and affine models and decompose the market price of Brownian risk into the volatility of the long bond plus an additional risk premium defining the permanent martingale component in the long-term factorization. C210, Evanston, IL, 60208, United States of America, LikuanQin2012@u.northwestern.edu, Vadim Linetsky

4 - An Inventory Control Model for Modal Split in Transportation: A Tailored Base-Surge Approach Chuanwen Dong, Kuehne Logistics University, Grosser Grassbrook 17, Hamburg, Germany, chuanwen.dong@the-klu.org, Sandra Transchel, Kai Hoberg We study an inventory control problem where a firm ships products from a plant to a distribution center via two transportation modes, a slow and cheap mode (e.g., rail) and fast but expensive mode (e.g., truck). In the slow mode, products are shipped with a constant volume, whereas fast-mode ordering follows a base- stock policy every period. We extend the tailored base-surge (TBS) policy to less frequent slow-mode shipments and present an approximated analytical solution approach. 5 - Optimal Time to Reposition Inventories in Multi-location Centralized Networks Repositioning of inventories between locations aims to decrease the impact of inventory imbalance caused by e.g. imperfect demand information or delayed delivery. In practice, it is often done via lateral transshipments. Our dynamic transshipment policy answers questions – when within the order cycle, how much, and from which location to transship to maximize the revenue of the network. A myopic policy and a near-optimal policy based on approximation are suggested for real-size problems. WA04 04-Room 304, Marriott Economics II Contributed Chair: Maryam Razeghian, Doctoral Student, EPFL, CDM-ODY 4.16, Station 5, Lausanne, VD, 1015, Switzerland, maryam.razeghian@epfl.ch 1 - Discrete Choice Modeling Approach to Decide The Digital Divide Policy Issue Subhabrata Bapi Sen, Adjunct Faculty, Sillberman College of Business, 32 Rolling Hill Dr, Chatham, NJ, 07928, United States of America, bapi45@fdu.edu Discrete choice framework to address the digital divide issue as reported in NY Times story “F.C.C. Chief Seeks Broadband Plan To Aid The Poor” Bridging Digital Divide. To focus on the main issue - we need to access to Broadband demand using a conditional logit formulation. The service attributes/demographics will be explanatory variables. June 6, 2015 in The Wall Street Journal article “Is the U.S. Ready to pay for ‘Quad Play’? is analyzed here. 2 - On the Relevance of Probability Distortions in the Extended Warranties Market Mike Abito, Assistant Professor, University of Pennsylvania (Wharton), 3620 Locust Walk, SHDH 1407, Philadelphia, PA, 19104, United States of America, abito@wharton.upenn.edu We study the reasons for high profits in the extended warranties market. Using data from a big US consumer electronics retailer, we find that overweighting of failure probabilities is a relevant factor in determining economic outcomes: without probability overweighting, profits drop by 90% and consumer surplus more than doubles. We also find that overweighting is affected by the environment and is reduced with learning. 3 - To Share or Not to Share: Adjustment Dynamics in Sharing Markets Maryam Razeghian, Doctoral Student, EPFL, CDM-ODY 4.16, To describe and further predict the growth of sharing markets, we construct a model for the dynamic sharing decisions of heterogeneous suppliers in a market with frictions, allowing for a mismatch between supply and demand. In each time period, an agent can enter or leave the sharing market, subject to an adjustment cost. We provide a closed-form expression for the nonlinear evolution of the rational-expectations equilibrium in this economy, typically resulting in an S- shaped diffusion pattern. Station 5, Lausanne, VD, 1015, Switzerland, maryam.razeghian@epfl.ch, Thomas Weber Olga Rusyaeva, Kuehne Logistics University, Grosser Grasbrook 17, Hamburg, Germany, olga.rusyaeva@the-klu.org, Joern Meissner

375