Informs Annual Meeting 2017

TB26

INFORMS Houston – 2017

TB27

4 - An Aggragator Business Model for Distribution Level Electricity Market Shmuel S. Oren, Professor, University of California at Berkeley, 57 Hill Rd, Berkeley, CA, 94720, United States, oren@ieor.berkeley.edu, Clay Campaigne The Californa ISO has introduced a tariff that recognizes aggragators of distributed resources at the distribution level as wholesale market participants. In this talk I will describe an end to end business model in which aggragators bundle demand response at the distribution level with intermitten renewable energy resources so as to create a firm energy product that is offered into the wholesale market. We describe the contract structure and incentive mechanism offered to customers and the deployment and bidding strategy of the bundled wholesale product. 5 - Modeling of Operating Reserve Demand Curve in ERCOT Electricity Market Hailong Hui, ERCOT, Taylor, TX, United States, hhui@ercot.com ERCOT has implemented the Operating Reserve Demand Curve (ORDC) in its Real-Time market since June 1, 2014. The Real-Time prices determined by Security Constraint Economic Dispatch (SCED) would be increased by the Real- Time Reserve Price which is determined based on remaining reserves in the system and a predefined ORDC to reflect the incremental value of scarce operating reserves. This presentation will discuss the current implementation of ORDC and the impact since it went live. Future possible improvement for ORDC in ERCOT market will also be discussed. 350B Clustering and Social Network Analysis Sponsored: Telecommunications Sponsored Session Chair: Eli Olinick, Southern Methodist University, Dallas, TX, 75275-0123, United States, olinick@lyle.smu.edu 1 - The One Time Period Least Cost Influence Maximization Problem on Social Networks Rui Zhang, University of Colorado, Boulder, CO, United States, Rui.Zhang@Colorado.edu, S. Raghavan We consider an influence maximization problem where there is requirement that the diffusion of influence takes place within one time period, which is referred to as the one time period least cost influence problem (1TPLCIP). First, we propose a linear-time algorithm for the 1TPLCIP on trees. More importantly, we present a tight and compact extended formulation for the 1TPLCIP on trees. Then, we project the extended formulation onto the space of the original payment and node selection variables. Finally, we observe that the good formulation for trees is also a valid formulation on general graphs resulting in a stronger formulation for the 1TPLCIP on general graphs and discuss our experiment results. 2 - Finding the Hierarchy of Dense Subgraphs Ahmet Sariyuce, Sandia National Laboratories, Livermore, CA, United States, asariyu@sandia.gov Finding dense substructures in a network is a fundamental operation, with applications in social networks, visualization, and cyber-security, to name a few. Current dense subgraph finding algorithms usually optimize some objective, and only find a few such subgraphs without providing any structural relations or distribution over the graph. We define the nucleus decomposition of a graph, which represents the graph as a tree of subgraphs, called nuclei, where smaller cliques are present in many larger cliques. The tree of nuclei consistently gives a global, hierarchical snapshot of dense subgraphs which are higher quality than the state-of-the-art. 3 - Analysis and Comparison of Node Centrality Measures Eli Olinick, Southern Methodist University, Dept Eng Mgmt Info and Systems, P.O. Box 750123, Dallas, TX, 75275-0123, United States, olinick@lyle.smu.edu We propose a new clustering algorithm based on hierarchical maximum concurrent flow (HMCF) and its duality relation to the sequence of sparsest cuts, and discuss theoretical properties which make it more accurate and often more robust than many popular algorithms in the literature. We present a new measure of node centrality, determined from the HMCF, called flowthrough centrality, and empirical results comparing its improved stability relative to currently used centrality measurements employed in social network analysis when knowledge of the network topology is incomplete or in transition. TB26

350C Psychology Neural Cognitive Computing Invited: Social Media Analytics Invited Session

Chair: Jiayin Qi, Shanghai University of International Business and Economics, Room 338, Bocui building, 1900th Wenxiang Road, Songjiang District, Shanghai, China, Shanghai, 201620, China, qijiayin@139.com 1 - Braking and Snowballing: The Moderating Roles of Adaptive and Maladaptive Cognitive Emotion Regulation Strategies between Job Stressor and General Health Charles Deng, PhD, Shanghai University of International Business and Economics, Shanghai, 201620, China, dengsc@live.com The present study investigated the moderating role of cognitive emotion regulation strategies between job stressor and general health. With the survey data from 495 employees, we found that job stressor was significantly and negatively associated with general health in both males and females with demographics being controlled. 2 - Recognition, Retrieval and Content Analysis for Massive Speech Data Ji Wu, PhD, Associate Professor, Tsinghua University, Beijing, 100084, China, wuji_ee@tsinghua.edu.cn, Xiangling Fu In this talk, I will talk about a speech content analysis platform developed at the Multimedia Signal and Intelligence Information Processing Lab at Tsinghua University and iFlytek for call centers in China that integrates ASR, audio indexing and spoken language understanding techniques. Quality inspection of custom service can be automatically achieved through this platform and its analysis results will be of a great help to strategy design and initiative marketing. The platform has been widely used in call centers in twenty provinces in China. 3 - Experimental Techniques for a Novel Neural-network-like System for Image Classification Lai Xue, Shanghai Xiayun Culture Communication Co., Ltd, Shanghai, China, sheley1998@gmail.com Artificial neural networks are extremely robust in areas such as image classification and speech recognition due to the fact that they could be trained instead of having to be explicitly programmed. This research presents several experimental non-biologically-inspired techniques for a neural-network-like system. An open-source web based implementation of this system is constructed to evaluate the efficiency of these techniques, provide realtime visualizations, as well as provide an interface for user interaction in order to further experiment with the capability of said system. 4 - Diffusion of Collective Action through Social Media Networks Rostyslav Korolov, Doctoral Student, Rensselaer Polytechnic Institute, Troy, NY, 12210, United States, korolr@rpi.edu, Molly Renaud, David Mendonca, William A. Wallace Previous work has shown that social media communications can be used to predict collective responses to certain events in the real world. However, the actual processes that enable this prediction remain unexplained. Considering collective behavior as a diffusion process, we expect the dynamic structural properties of the social network to play a role in the diffusion of collective behavior. We explore the possibility of using structure and dynamics of networks found in social media to predict collective action and identify events that may trigger it.

312