Informs Annual Meeting 2017

MC69

INFORMS Houston – 2017

3 - Self-segmented Classification via Adaptive Network Lasso with Application in Purchase Prediction Shuyu Chu, IBM T.J. Watson Research Center, Yorktown Heights, Yorktown Heights, NY, 10598, United States, cshuyu@vt.edu, Huijing Jiang, Xinwei Deng, Zhengliang Xue, Naoki Abe Predicting the purchase likelihood of potential clients is essential in finding the optimal pricing strategy in business. However, the heterogeneity within clients and products lead to different purchase behaviors. Thus, building one global model on all data is no longer appropriate. In this work, we created a self- segmented classification method to perform segmentation and model fitting simultaneously. This method ensures that the estimated models under different segments are distinctive and data points with a common model structure are grouped into the same segment. The performance of the method and its merits are illustrated by numerical examples and a case study with the IBM pricing data. 4 - A Spatial Statistical Method for Designing 3D Laser Scan Paths Romina Dastoorian, Western Michigan University, Kalamazoo, MI, United States, romina.dastoorian@wmich.edu, Lee Wells The emergence of 3D laser scanning technologies has significantly enhanced inspection system capabilities by providing high-density point cloud data. However, these new capabilities also bring new challenges regarding both the effective and efficient use of point cloud data. To improve the analytical power these datasets provide, it is essential to reduce or eliminate measurement noise. This presentation will introduce a methodology for modeling scan-path induced measurement. This methodology is based upon a spatial statistics model that assumes a point cloud’s spatial correlation is the combined effect of both surface induced and scan-path induced is variation. 371D Optimization Models in Environment and Sustainability Sponsored: Energy, Natural Res & the Environment Environment & Sustainability Sponsored Session Chair: Hadi Karimi, Clemson University, hkarimi@g.clemson.edu 1 - Grow your Own Feedstock to Mitigate the Impact of Biomass Supply Uncertainty: A Case of Cellulosic Biofuel Investment Yihua Li, Graduate Student, Iowa State University, 230 Raphael Ave, Unit 12, Ames, IA, 50014, United States, yihuali@iastate.edu, Guiping Hu A decision-maker wants to enter the cellulosic biofuel market, where his primary option is to invest in a biofuel facility. Corn stover can serve as a main feedstock, while its yield fluctuates due to change in weather conditions. To hedge against supply uncertainty, he has a secondary option of investing in a land to grow his own feedstock (e.g., switchgrass) to achieve dual sourcing in biomass supply. The optimal investment timing for both investments are investigated via a real options approach. The effect of facility and land lead times, as well as risk aversion, are also discussed 2 - Meeting Corporate Renewable Power Targets Danial Mohseni-Taheri, TA Coordinator, University of Illinois, Chicago, IL, 60607, United States, smohse3@uic.edu, Selvaprabu Nadarajah We study the procurement problem faced by a firm trying to purchase a target percentage of its power demand from renewable sources by a future date and sustain this target going forward. We provide analytical insights into the structure of the procurement portfolio when employing recent variants of power purchase agreements. In addition, we discuss the challenges of setting up such a portfolio over a long time horizon and provide computational solutions. 3 - The Optimal Operation Policy of a Grid-connected Microgrid MC69

4 - A Stochastic Multiobjective Optimization Model to Analyze the Economic and Environmental Impacts of Biopower Supply Chains Hadi Karimi, Clemson University, 1108 Tiger Blvd., Apt. 117, Clemson, SC, 29631, United States, hkarimi@clemson.edu, Sandra D. Eksioglu, Michael Carbajales-Dale In this study we propose a stochastic multiobjective optimization model to analyze the economic and environmental impacts of biomass power supply chains. While generating electricity from biomass is known to provide an economically attractive option for coal-fired power plants to comply with emission regulations, the total footprint of the associated supply chain still needs to be carefully investigated. The multiobjective model aims at maximizing the potential profits and minimizing the life cycle emissions in the biomass supply chain. We consider the uncertainty in biomass feedstock quality and use sampling-based chance constraints to capture the uncertainties in the optimization model. 371E Data Mining Contributed Session Chair: Matthew Lanham, Purdue University, West Lafayette, IN, United States, lanhamm@purdue.edu 1 - Graph Regularization Based Framework for Product Filtering in E-commerce Mukul Gupta, Indian Institute of Management Indore, Indore, 453331, India, mukulg@iimidr.ac.in, Rajhans Mishra The multitude of choices for users of e-commerce has created a challenging problem for them to select the product of their choice in a reasonable time. Product filtering for helping a user to narrow down to a product of their choice is getting the tremendous attention of researchers. In this work, we propose a preference diffusion approach using graph regularized framework for product filtering. The user-product network with the attribute information of products is formed to diffuse the preference of a user to generate recommendation after filtering of less relevant products for that user. 2 - A Cognitive Framework for Optimal Service Composition MC70

Ahmed Nazeem, Research Staff Member, IBM.Research, IBM.Almaden Research Center, San Jose, CA, 95123, United States, nazeem.35@gmail.com

With the evolution of the cloud based services and the Internet of things, there is an increasing need to develop an end-to-end framework that automates the solution design. 3 - Recommendation System Incorporating User Profile and Sequential Information

Rajhans Mishra, Indian Institute of Management-Indore, Faculty Block B, First Floor, Rau, Indore, 453331, India, rajhans111@gmail.com, Mukul Gupta

Recommendation system generates suggestions to the users as per their preferences with respect to various products and services. The proposed work focuses on generation of recommendations using user profile, content and sequential information with respect to user activities. User profiling may be done on various user specific parameters while sequential and content information can be captured through user activities as web navigation, product /service selection etc. Considering user profile, content, and sequential information collectively may help to capture the preferences of the users. 4 - A Text Based Decision Support System to Detect Fraudulent Bank Activity lanhamm@purdue.edu, Jeffery Mei, Alan Abrahams, Rich Gruss We develop a text analytics decision support system to detect fraudulent bank account activity. This research is motivated by the 2016 Wells Fargo scandal which entailed a $190 million settlement with federal regulators and prosecutors for opening more than two million deposit and credit card accounts without customer authorization. We show that our DSS can be used by regulators and financial institutions for internal control. Matthew A. Lanham, Clinical Assistant Professor, Purdue University, West Lafayette, IN, United States,

Jong-hyun Ryu, Hongik University, 2639 Sejong-ro, Jochiwon-eup,, Sejong, 30016, Korea, Republic of, jongh.ryu@gmail.com, Dong Gu Choi, Daiki Min

This work considers the optimal operation of a gird-connected microgrid which comprises renewable energy resources, energy storage devices, controllable loads, and controllable generators. To satisfy the increasing electricity demand in a sustainable way, the efficient use of renewable energy is needed and the microgrid concept is one of the approaches. In this work, the optimal operating policy to minimize the generation cost in the microgrid is identified considering the uncertainty in renewable energy generation and load demand.

228