Monday Posters: 2:45 - 3:35 | Grand A, B, C and Foyer
Past research has documented the influence of low cost carriers on the performance of full service carriers and vice versa. This study intends to add regional carriers to the comparison, since this category of airlines has grown substantially in the past decade. The proposed model separately examines the operational and financial efficiencies of the three categories of airlines. Beyond the 18 Brazilian airlines considered, we also include a virtual airline, with the best values of inputs and outputs; in other words, this virtual airline has the best of each variable considered in both categories, operational and financial. This insertion allows determining the potential improvements even for the best airlines, which without the virtual airline, were the benchmarking. Empirical results identified that full service and low cost airlines have better operational efficiency than financial efficiency, and for the regional carriers the opposite results were found; that is, the financial efficiency is superior to the operational efficiency.
Maria Christina Gramani is an Associate Professor at Insper – Institute for Education and Research (São Paulo, Brazil), working mostly on two research areas: Operations Research and Operations Strategy. The first one includes the resolution of optimization problems such as production planning, routing/transportation, cutting stock, and assignment problems, among others. The second area emphasizes the efficiency/performance analysis in operations and non-profit sectors, including education and healthcare systems. I have been published in academic journals such as the European Journal of Operational Research, International Journal of Production Economics, Expert Systems with Applications, International Transactions in Operational Research, Plos One, as well as in other Brazilian journals.
The problem of assigning optimal tolls to the arcs of a multi-commodity transportation network can be formulated as a bilevel mathematical program. The problem can be interpreted as finding a trade-off among tolls generating high revenues and those being yet attractive for the users.
The aim of the poster is to describe an algorithm based on the allowable ranges to stay optimal (ARSO) resulting from sensitivity analysis after solving the lower level problem. With this powerful tool, one can analyze possible changes in the coefficients of some variables in the objective function within these allowed ranges without affect the optimal solution.
In addition, the proposed technique also uses the concept of “filled function”, which is applied under the assumption that the local maximum (in our case) has been found. Then the “filled function” technique helps one find another local maximum, better than the previous ones, or determine when to stop.
Vyacheslav V. Kalashnikov obtained his PhD in Operations Research (OR) in 1981 from the Institute of Mathematics of the Siberian Division of the USSR Academy of Sciences in Novosibirsk. He was awarded his Dr. Sc. (Habilitation Degree) in OR in 1995 from the Central Economics and Mathematics Institute (CEMI) of the Russian Academy of Sciences in Moscow, Russia. He is the author and co-author of 4 monographs and more than 60 papers, published in many prestigious journals and publishing houses. He has advised 10 Ph.D. students and 25 master students in universities in Russia, Mexico, and Ukraine.
We present a working and in-use mathematical model of the sales forecasting, business operations and strategic planning for products used to detect seasonal infectious diseases. BioFire Diagnostics designs, manufactures and sells human diagnostic tests for infectious diseases, most notably, a test that detects and differentiates flu-like respiratory illnesses. Our tests have a short shelf life and the demand is seasonal with sales during the flu season vastly exceeding baseline demand. BioFire’s business operates on the ‘razor-razorblade’ model; tests are run in a durable instrument, and each patient test is performed in a single use medical device. Our install base is doubling in less than 2 years. The mathematical models use combinations of techniques including dynamical systems, statistics and curve fitting, Monte Carlo simulation and optimization to account for sales processes, customer behavior, operations, and provide management with a hypothesis testing tool used for business planning. We will describe details of how the model incorporates CDC infectious disease surveillance data to estimate characteristic behavior of the flu season and how those data are used plan capacity, supplies and finished goods inventory levels.
Kirk Smith has a diverse academic background and is always eager to learn … self motivated … honest … committed. I try to always be above average in anything that I care about.
The settlement of corporate debts may be influenced by the interconnectedness of institutions in the debt network. An innovative algorithm is developed to liquidate corporate debt network and improve the efficiency of debt settlement. This algorithm avoids the massive search for debt circles and debt chains by replicating debt network with forward debt chain and backward debt chain and can reduce the aggregate debt amount to the minimum in polynomial time. In the age of big data, the algorithm
becomes especially efficient for clearing the large scale debt network based on the mass of corporate debt data. The algorithm is demonstrated through a five-firm case study and a large-scale sovereign debt analysis using real debt data from ten developed countries. Intermediate debt relationships are removed and debt relations in the debt network become simple and straightforward. Thus, institutions are released from heavy debt loads.
Shuzhen Chen is a doctoral student at the School of Management, University of Science & Technology of China. She is a key member in a National Science & Technology Pillar Program during China’s twelfth five-year plan period. She has designed an algorithm to optimizing corporate debt network, which is one major achievement of the program. Her research focuses on decision analysis, algorithm design and risk management.
5 - Application of O.R. to Customer Service, Emergency Response and Marketing in a Gas Company: Successful and Failure Cases
Tokyo gas which is Japan’s largest city gas supplier has been applying operations research to customer service, emergency response and marketing for 40 years. For example, we have done development of statistical human resources planning in a call center, optimization of dispatching the employees in case of a huge earthquake, and area marketing for gas appliance. We focus on how to develop feasible solutions in real business, how to find applicable topics and how to organize O.R. projects. Factors which determine projects’ success or failure are not sophistication and novelty of statistical methods but project organization, priority to maintenance of model, cooperation with outside experts and application of IT.
Toshinori Sasaya is a Senior Researcher for Tokyo Gas. He has conducted various O.R. projects (supply chain management, emergency response, customer service, product development and marketing) in the O.R. team of Tokyo Gas. Tokyo Gas won the Practice Award in 2014 in the Operations Research Society of Japan. He earned his Bachelor’s and Master’s degrees in Urban Engineering from the University of Tokyo.
The U.S. Marine Corps (USMC) Logistics Command Studies and Analysis Division created ROME to evaluate repairable equipment and directly influence budget planning decisions. The USMC, fighting a global war to support and protect U.S. interests, has seven million ground equipment items and needs to rationalize annual expenditures in support of depot maintenance planning and execution. For example in FY17, the projected depot maintenance budget is $229M versus a requirement of $315M. In a growing fiscally uncertain environment, only the most mission critical equipment gets the highest priority for repair. ROME provides a solution to a problem that, once hard to quantify, is solved in less than a minute, and supports an Enterprise process that historically was solved using spreadsheets and stubby pencils. ROME is a mixed integer program implemented in the GAMS software package and solved using the IBM ILOG CPLEX optimizer.
Mr. Norm Reitter, the Director of Analytics for CANA Advisors, has over 21 years of experience in logistics and applying operations research-based methods. He received an MS in Operations Research from NPS in 1993.
Recently, many studies on public transportation systems using the smart card big data have been conducted to provide the public with better services such as traffic congestion analysis, arrival time estimation and transit route planning. In several literatures, service qualities of public transportation systems were assessed via user survey or limited sample transaction data due to difficulty of extracting entire transaction data. However, more detailed transaction data can currently be collected and analyzed because of the advancement of smart card systems and large-scale data processing technology. In this research, a public transit convenience index is presented which can be calculated by using travelers’ transaction data including movement distance, travel time, and transfer frequency of travelers. The convenience index is applied to and verified with a smart card transaction data of Seoul metropolitan city. The proposed index can be used to evaluate and improve transit routes between regions, intervals of transportation means, and travel convenience in the viewpoint of travelers based on more realistic transaction data.
Kyuhyup Oh received a BS from the Department of Industrial and Management Systems Engineering at Kyung Hee University, Yongin, Republic of Korea, in 2010. He is currently a PhD candidate in the same department. His research interests are big data analysis and process mining.
8 - Discovery of Information Diffusion Process Based on Bayesian Belief Networks in Social Network Services
In online social network services (SNS) such as Twitter and Facebook, information is shared and propagated rapidly among users. To understand such information flows in SNS, we extract event logs from the SNS though open API and then discover information diffusion processes by using process mining techniques. While typical existing algorithms for process discovery deals with business events in enterprise information systems, we introduce how to discover process models from event logs on social media. In this research, an information diffusion process is constructed from the event log in Twitter which includes the traces of friends’ responses (e.g. retweets and replies) for each tweet. To resolve it, a new process mining technique based on Bayesian belief networks is also presented to represent probabilistic process models. As the growth of SNS, the proposed methodology can be effectively applied to many fields such as viral marketing and advertisement on the Web.
Jae-Yoon Jung received BS, MS, and PhD from the Department of Industrial Engineering at Seoul National University, Seoul, Republic of Korea, in 1999, 2001, and 2005, respectively. He is currently an associate professor in the Department of Industrial and Management Systems Engineering at Kyung Hee University, Seoul, Republic of Korea. His research interests include business process management, business analytics, and big data analytics.
Higher earnings quality provides more information about the firm’s financial performance. Thus accurately measuring earnings quality is highly important to standard setters, auditors, regulators, analysts, investors, as well as accounting and finance researchers and educators. Although earnings quality has received considerable attention from researchers, there is still no consensus on measuring or qualifying it. The residual from regression is widely used overall qualification of whether a stock exchange has low or high earnings quality. We propose the use of optimistic bias as an overall qualification of earnings quality. Optimistic bias occurs when actual accruals exceed predicted accruals from multiple predictive models; in this case, a linear regression and an artificial neural network. Optimistic bias is applicable to any company or industry type.
Adam Fadlalla is a Professor of Information Systems in the Department of Accounting and Information Systems of Qatar University. He holds an MBA from Miami of Ohio, an MSc in Computer Science and a PhD in Information Systems from the University of Cincinnati. His research spans both the conceptual and the applied aspects of Information Systems, and his current research interests include modeling to improve organizational outcomes.
In airline revenue management, an accurate prediction of cancellations is crucial, since a significant number of bookings are cancelled before departure. Accurate estimation of cancellation behavior is essential for airlines, so that they can allow more reservations on a flight than there is physical capacity (“overbooking”), which is a significant source of revenue. Cancellation probabilities depend in a complex manner on several flight-related and passenger-related attributes. A proportional hazard model is applied to predict the “hazard rates”, i.e. the conditional risks of a reservation being cancelled. These are used to forecast the expected number of cancellations depending on bookings on hand and forecasted future demand. We enhanced the standard maximum-likelihood estimator in order to obtain practicable processing time and memory consumption. The new method provides stable predictions and improves accuracy significantly compared to a time series approach.
Heiko Schmitz received his PhD in computer physics from Max-Planck Institute for Polymer Research/University of Mainz. He participated in the optimization of cell phone networks at Siemens and has been working as an expert for airline revenue management at Lufthansa Systems for seven years.
Developing a robust supply chain responsive to changes in supply and demand requires a deep understanding of inventory policies and the implications of shifts in inventory management to service level and cost. Manufacturing and assembly flows include numerous vendor and manufacturing nodes and involve many process steps and inventory points. We developed a discrete event simulation to analyze demand variability as well as a number of different inventory policies. The simulation provides greater insight into the production flow as well as interconnected inventory locations because it includes manufacturing, assembly, and inventory sites. Using the simulation, we can test a variety of inventory strategies by selecting different replenishment policies and evaluating their impact on customer service level and holding cost. In addition, we have the ability to run the simulation against alternate demand patterns, which provides an enhanced capability to further analyze service level and cost implications under different circumstances
Jennifer Schilling is an Operations Research Engineer in the Supply Chain Intelligence and Analytics group at Intel Corporation. She holds a BS in Applied Mathematics with a minor in Computer Science and an MS in Computational Operations Research, both from the College of William and Mary. In her role at Intel, she provides high impact analytics and tools using simulation, optimization, and data analysis to enable supply chain success and excellence.
We developed a new fuzzy ranking model for using in searching algorithm based on the uncertainty in the center of the searched location. In an area with diverse neighborhoods, the searched point is not necessarily close to all neighborhoods. Therefore, the points close to one of the neighborhoods do not have priority in returning list. Our proposed model by using general type-2 fuzzy membership value returns the listing based on the closeness to at least one of the neighborhood and also the size of neighborhoods. General type-2 fuzzy membership value of each point to each neighborhood is defined by the primary variable which indicates the closeness to the center of the neighborhood; and the secondary variable which indicated the position of the neighborhood compared to other neighborhoods in the searched area. The quality of ranking points is increased by considering both primary and secondary variable at the same time.
Samira Malek Mohamadi Golsefid is a PhD candidate in Industrial Engineering at Amirkabir University of Technology, Tehran, Iran. Her research is focused on fuzzy overlapping community detection model for complex social networks. She also revised the BASc degree in chemical engineering from Islamic Azad University of Science and Research of Tehran, Tehran, Iran, and the MASc degree in information technology and e-commerce engineering from Iran University of Science and Technology Tehran, Iran. Her research interests include pattern recognition, data mining modeling and applications, fuzzy sets and systems, knowledge-based expert systems, social network analysis.
In my role at SAS I routinely speak with executives who tell me that the ability to explain analytical results is just as important as producing them. In this poster I will share the top five tips and tricks I have learned in my several years of presenting complicated analytical results to audiences, both technical and lay. At the root of successful “selling” of an idea is an understanding of the objective functions of the audience. That is, why do they care about what you have to say? What is the opportunity cost of inaction? My poster will include tricks and tips for understanding the audience’s professional and educational background. Other tips we will discuss are more technical in nature, involving the way to couch various alternatives and how to present these solutions so as to increase the likelihood of adoption.
Kenneth Sanford is a Senior Research Statistician in Advanced Analytics R&D and an econometric evangelist at SAS. He is responsible for helping to integrate the advanced analytical procedures being created by the development team into SAS’ business solutions. Sanford joined SAS after working for a large private consulting firm and holding faculty positions at the University of Cincinnati and Middle Tennessee State University. His work on optimal pricing has been published in the Southern Economic Journal and the Journal of Sports Economics. He holds a PhD in economics from the University of Kentucky.
This research provided a data driven approach to predict a geographic area’s expected recruit production for a given month. The authors served as the lead and senior analyst for the project, developing the approach, conducting the analysis and communicating the results of the analysis to senior decision makers. The model will allow leaders of the military recruitment enterprise to more efficiently allocate their resources and identify an area’s potential compared to current best practices. The general methodology employed could be applicable to other industries where the need exists to identify an areas potential in order to set production goals and allocate marketing and sales resources. The predictive model can be used directly, but has also been incorporated into a web-based game intended to train leaders on the use of analytics to enable decision making.
Jonathan K. Alt is a U.S. Army operations research analyst currently serving as an Assistant Professor in the Department of Operations Research at the Naval Postgraduate School. Lieutenant Colonel (LTC) Alt holds a Ph.D. in modeling and simulation and an MS in operations analysis from the Naval Postgraduate School, a M.Ed. from the University of Georgia, and a bachelor’s degree from the United States Military Academy. LTC Alt served in a variety of infantry assignments prior to becoming an operations research analyst, where he has served in both analytic agencies and as a brigade combat team analyst in Iraq. Most recently, LTC Alt served as the director of Training and Doctrine Command Analysis Center-Monterey, an applied research and development organization, and a member of the board of directors of the Training and Doctrine Command Analysis Center. He serves as an associate editor for the Journal of Defense Modeling and Simulation and is a member of the Military Operations Research Society and the Institute for Operations Research and Management Science.
The emergence of connected vehicles technology creates game changing opportunities for vehicle service and maintenance. The ability to real-time capture the operational and environmental data from each vehicle combined with the periodic direct measurement of the actual wear on only a subset of vehicles (in case of a warranty repair or during a dealer checkup) allows to accurately and continuously predicts the amount of wear on the entire fleet. Wear prognostics model is implemented using a variation of the Kaplan-Meier Product Limit Estimator and a modified survival regression model. This prediction can be conveyed to a driver and to a manufacturer or a dealer so that appropriate repair/maintenance action can be taken long before the system passes a critical wear point. Furthermore, these prediction models can also be used to determine, for each vehicle, the projected incremental mileage or number of days remaining to a critical wear point thus enabling to optimize a schedule of service appointments. The models are illustrated through application of brake pad wear prognostics.
Dr. Oleg Gusikhin is a Technical Leader at Ford Research and Advanced Engineering. He received his Ph.D. from the St. Petersburg Institute of Informatics and Automation of the Russian Academy of Sciences, MS in Electrical Engineering from St. Petersburg State Technical University, and an MBA from the Ross Business School at the University of Michigan. For over 20 years, he has been working at Ford Motor Company in different functional areas including Information Technology, Advanced Electronics Manufacturing, and Research & Advanced Engineering. During his tenure at Ford, Dr. Gusikhin has been involved in the design and implementation of advanced information technology and intelligent controls for manufacturing and vehicle systems. Dr. Gusikhin is a recipient of 2014 INFORMS Daniel H. Wagner Prize, 2009 Institute of Industrial Engineers Transactions Best Application Paper Prize in Scheduling and Logistics, and two Henry Ford Technology Awards. He is an Industry Vice-Chair of IFAC Technical Committee “Manufacturing Modeling for Management and Control,” and a Lecturer in the Industrial and Operations Engineering department at the University of Michigan.
A common scenario in Business Analytics is to identify the relationship and influence of one or few dimensions/attributes with the target of interest and the goal is to identify a series of tabular reports that illustrate the important target-dimension relationships. However, for datasets with a large number of dimensions, it becomes prohibitive to generate and analyze all possible tabular reports. Thus we propose an automatic, structured and scalable search process to help business users achieve the goal and also provide some valuable plain-language insights to them.
Jane Chu is a senior statistician for IBM Business Analytics. She holds a PhD and MBA in Statistics and Econometrics from Booth School of Business, University of Chicago. She works with researchers across different IBM teams to develop new data mining and statistical analytics for big data. Some recent developed analytics include relationship discovery, temporal causal modeling, and spatio-temporal exploration and modeling and many of them have been included into IBM Watson Analytics. She has filed 7 patent applications and 5 papers in the first 5 years of her tenure in IBM.
This study proposes game models for different resource allocation strategies that focus on the impact of capability exploration and exploitation on attaining core competence and competitive advantage. Having more efficient capabilities generates distinctive competence which is the source of competitive advantage and leads to superior profitability. The study explores the effect of the two opposing strategies (capability exploration and exploitation) on product value creation based on the resource-based view, and investigates the influence between rivals’ efficiencies of capabilities by making use of game theory. Four game models are proposed under four possible competitive scenarios according to the VRIO framework. The models will enable firms to identify their competitive positions and to select an optimal resource allocation strategy for enhancing or sustaining their competitive advantage under various competitive scenarios.
Chinho Lin is a chair professor of the Department of Industrial and Information Management & Institute of Information Management at National Cheng Kung University, Taiwan, (ROC). He received his PhD in Business Administration from the City University of New York. His works have been published in European Journal of Operations Research, Omega, Journal of Operational Research Society, Information & Management, Decision Support Systems, Decision Sciences, International Journal of Production Research, and other journals. His current research interests include knowledge management, supply chain management, quality and reliability management, and technology management.
BIAS Optimizer is a computer based BI instrument to address business management problems in various industries. The supply-demand-cost configuration is the major paradigm addressed by BIAS Optimizer. This configuration is often encountered in transportation and logistics environment. BIAS Optimizer is aimed at solution of various optimization problems such as asset allocation, chain supply, power distribution and others. BIAS Optimizer platform encompasses optimization algorithms in conjunction with specially developed data structure allowing efficient solution of large scale optimization problems.
BIAS Optimizer data structure is flexible and allows for easy integration with virtually unlimited variety of business specific data resources. The developed data structure enables smooth formalization of business rules involved in particular business.
BIAS Optimizer has been successfully implemented as a business operation tool of a major railcar owner (over 30,000 rail cars).
Ilya Buzytsky leads the Big Data Solutions and Technology and Infrastructure teams, focusing on the creation of Big Data pipelines, BI solutions, and analytical platforms. Ilya has extensive experience building big data solutions in both Microsoft and open source environments. Prior to joining Society Consulting and forming BIAS Intelligence, Ilya spent 10 years at Microsoft, and 10 years as an entrepreneur and consultant focused on solving business problems through technology.
The PuzzlOR is a bi-monthly column published in OR/MS Today and Analytics magazines that has been running since 2008. Each article consists of a simple puzzle that requires Operations Research techniques to solve. A winner, chosen randomly from all correct entries, receives a prize.
Explaining the complex concepts of Operations Research to lay audiences remains one of the biggest challenges that our field faces for widespread acceptance of O.R. in business and industry. The PuzzlOR column helps with this problem by breaking down the fundamental O.R. methodologies (optimization, simulation, decision trees, predictive modeling) into simple questions that anyone can understand. In doing so, the PuzzlOR introduces the concepts of O.R. in a fun and simple way to people who have had little exposure to it.
John Toczek is the Sr. Director of Decision Support & Analytics at Aramark in Philadelphia, PA. He has a background in both engineering and operations research. Mr. Toczek has extensive experience in designing and developing decision support systems for dynamic business environments and has had over eleven years of experience helping clients make better, more informed decisions by championing the use of data and advanced O.R. techniques. Mr. Toczek earned his MS in Operations Research from Virginia Commonwealth University in 2005 and his BS in Chemical Engineering from Drexel University in 1996.
20 - Implementing the Affordable Care Act - Predicting the Operational Impact of Medicaid Expansion in San Diego County
The Affordable Care Act expands Medicaid eligibility to individuals with incomes up to 138% of the Federal Poverty Level. Medicaid expansion presented questions that the County of San Diego Health and Human Services Agency needed answered to prepare for the operational challenges presented by ACA: How many newly eligible individuals will there be? How many newly eligible individuals are uninsured and likely to apply for Medicaid? Where do these individuals live and how will they impact existing service centers? Where should we locate new facilities and/or staff to meet the demand generated by Medicaid expansion?
Using American Community Survey data, ArcGIS, and geocoded customer addresses, we answered these questions and identified sites for two new Medicaid eligibility service centers. We continue to use ACS data, together with incoming Covered California Medicaid application data, to estimate the remainder of the uninsured, Medicaid eligible population and to study variation in enrollment rates.
Ricardo Gutierrez is an Operations Research Analyst with the County of San Diego Health and Human Services Agency. His analyses inform decision that impact child welfare, public health, behavioral health, and social service outcomes for 3.2 million county residents. His recent projects include spatial analyses of customer data to identify optimal locations for service centers. Prior to joining the County of San Diego, he was an analyst with the Los Angeles Department of Transportation, where he completed cost-benefit analyses for capital projects. Ricardo received his BA in economics from Berkeley and a master’s degree in policy analysis from UCLA.
Given the widespread use and continuing development and popularity of analytic methodologies, INFORMS leadership determined a need for a personnel qualification standard. Thus was born the Certified Analytics Professional (CAP®) program. Those earning the CAP designation will have verified education, experience, ethics, effectiveness and successful completion of an examination. The credential is based on an analysis of practice and thus every question on the exam and every requirement for eligibility is tied directly to what a professional analyst does. These seven areas of assessment mirror the analytics project development: Business problem framing, analytics problem framing, data, methodology selection, model building, deployment and lifecycle management.
Dr. Wehrle is the Manager of the Certified Analytics Professional (CAP) program. Dr. Wehrle has extensive experience in the certification field and is the former Chair of the Certification Networking Group; she serves on an ANSI accreditation committee for the Council for Food Protection and is an ANSI accredited program assessor for the ISO/IEC Standard 17024 for Personnel Certification. Prior to working at INFORMS, Dr. Wehrle worked with certification programs for financial managers, fire sprinkler designers, thermal sprayers and others. She is the author of a number of published articles on the topic of certification and is a Certified Association Executive (CAE).
HIP is a rapidly growing cable television provider of health and beauty channels. Commercials and sponsorships managed by the HIP Ad Sales Team constitute the company’s major revenue source. In order to help the Ad Sales Team better manage advertising inventory, increase the total return on marketing investment (ROMI), and reduce the amount of manual labor required to create media plans, HIP wishes to develop a data-driven system that can generate media plans. The system will feature an integrated marketing platform that simplifies and automates the media plan generation process to improve the work efficiency and effectiveness of the HIP Ad Sales Team. Owl Research (OR) is proud to present this statement of work for Healthy Image Productions (HIP) to solve its Television Advertising Plan Optimization problem.
Qinlu (Louisa) Chen is a student in the University of British Columbia’s Master of Management in Operations Research program, graduating in December 2015. With an academic background in economics and finance, and project experience in marketing, she provides unique and insightful views on solving OR problems and effective methods for conducting data analysis, optimization, and simulation modeling. As an ambitious student with an outstanding academic performance and proven leadership skills, she looks forward to applying OR techniques to marketing or management consulting after she graduates.
Healthy Image Productions is a cable television provider of health and beauty channels that is rapidly growing with several channels already on the air and are continuously acquiring and developing channels. With its rapid growth, there is an increasing need to better manage inventory of unsold advertising shows to generate more revenue. In addition, ad planners manually generating media plan for each advertiser is very time consuming and labor intensive. An analytics program using a data-driven analytics approach will not only generate media plans that satisfy the advertiser’s requirements more efficiently, but also help users gain a better knowledge of inventory availability to make more effective marketing strategies
Jodie Lam is currently a candidate in Masters of Management in Operations Research at University of British Columbia, where she also received her Bachelor of Science majoring in Mathematics with a minor in Economics. She is on the committee of the UBC INFORMS Student Chapter and is also the winner in the SAS and INFORMS Analytics Section Student Analytical Scholar Competition.
We propose a methodology based on linear regression to identify inefficient (dominated) plans in the public health insurance exchanges (HIX) based on plans’ attributes. We quantify the excess premium that makes these plans appear non-competitive in our framework. Further, we discuss how to use our framework to assign letter grades to health payers based on the percentage of efficient (non-dominated) plans in their HIX portfolio. We illustrate our approach using HIX plan information in Pennsylvania and Massachusetts. Our methodology provides a novel way to analyze HIX offerings. Payers should better articulate the value proposition of plans labeled “inefficient” or “dominated”.
Dr. Aurelie Thiele is a Visiting Associate Professor at the MIT Sloan School of Management and an Associate Professor in the Department of Industrial and Systems Engineering at Lehigh University. Her research is on decision-making under high uncertainty with a focus on the financing of the healthcare system. She serves as the co-director of the Master of Science in Analytical Finance at Lehigh and is the Tutorials co-chair for the 2015 INFORMS Annual Meeting to be held in Philadelphia. Her research has been funded by the National Science Foundation, the CELDi consortium, an IBM Faculty Award and a Lehigh Innovation Grant, among others. She holds a “diplome d’ingenieur” from the Ecole des Mines de Paris and a MS and a PhD from MIT.
Tuesday Posters: 2:45 - 3:35 | Grand A, B, C and Foyer
After the fabrication process in semiconductor manufacturing, the wafer test, which involves evaluation of electrical properties, is conducted first to filter out defective dies in a wafer. Only the dies that are not defective proceed to the assembly and packaging steps in order to obtain packaged chips as final products. The chips are graded pass or fail through a functional test called the final test. An important issue is that several chips who passed the wafer test often fail the final test. This work is to predict the result of the final test for each semiconductor chip using the wafer test data before the assembly and packaging processes. We build a data-driven prediction model based on machine learning techniques. Our approach is promising in that accurate prediction of the final test can lead to the reduction of time and cost by eliminating assembly and the following steps for chips that are probable to fail.
Seokho Kang is a PhD candidate in the Department of Industrial Engineering, Seoul National University, Republic of Korea. His research interests include kernel-based learning algorithms, multiple classifier systems, dimensionality reduction, and their data mining applications to manufacturing systems. He has published a number of papers related to these areas in refereed journals and conference proceedings.
Tax-related identity theft is a rapidly growing crime that often imposes enormous financial, emotional, and time-consuming burdens on its victims. According to the Treasury Inspector General for Tax Administration there were almost 2 million suspected tax identity theft incidents in 2013, compared with about 440,000 in 2010. Tax Identity theft is a form of identity theft where the stolen identity is used primarily for filing false tax returns and obtaining refund, thereby delaying and impacting the legitimate tax payer’s ability to receive his or her tax refund. We describe highlights from two client engagements where we analyzed the key factors of tax identity theft and developed predictive models and analytically driven strategies to help solve their relevant business needs.
Sriram Tirunellayi is currently the Senior Director of Analytics, Identity and Fraud Data Sciences at Equifax Inc., where he is responsible for driving analytic product research and development and advanced analytical solutions that enable client value.
Sriram brings several years of broad experience developing and implementing Operations Research and Analytical solutions. He was previously with Barclays Bank for 4 years as a Manager in the Decision Science team, leading innovative analytical projects that involved building predictive models and optimization strategies for credit cards, merchant acquiring services and consumer loans pertaining to Risk, Acquisitions, ECM and Basel.
Prior to Barclays, Sriram spent 10 years in the airline industry working for Sabre Holdings in the areas of Pricing and Revenue Management for several domestic and international airlines. During that time his responsibilities included product development, product delivery and implementation and operations research support building optimization and forecasting models.
Sriram has a Master’s degree in Industrial Engineering from West Virginia University, Morgantown and a Bachelor’s in Mechanical Engineering from the National Institute of Technology, Surathkal, India.
Effective work teams are a key component of success in competitive environments. In his book Moneyball, Michael Lewis describes how Billy Beane used analytics to assemble an effective baseball team and triumph in the sports world. In this study, we extend the concepts of Moneyball to the business world and consider the factors that influence and contribute to sustained team effectiveness in high tech industries. Specifically, we identify critical variables and conditions that differentiate between high, average, and poor performing teams, and look for critical characteristics of short term and sustained team success. This work also includes considering assessment processes, building comprehension of which team practices are scalable, repeatable or unique, and designing and validating appropriate scoring methodologies. In addition, we consider the differences between personal and team success and how performance management reviews may influence team outcomes.
Dr. Genetha Anne Gray is a data scientist in the Talent Intelligence Analytics Organization at Intel Corporation where she analyzes talent supply chain management, career progression, and representation of women and underrepresented minorities. Before joining Intel, Genetha was a member of the technical staff at Sandia National Labs where she researched problems in the areas of systems engineering, the environment, security, and energy. She has a PhD in Computational & Applied Mathematics from Rice University and specializes in optimization under uncertainty, data fusion, model validation, and uncertainty quantification.
“Out of the Dark into the Analytical Light”: pushing a new start-up business unit through the analytical curve. In a fast-growing new business, how do you create business analytics to help predict needed resources when the historical data is sparse and of poor quality? How do you help an analytically naïve company or business unit become more analytically sophisticated in a relatively short amount of time? How do you help a newly purchased business unit integrate into existing company data structures? We look at creating metrics and key performance indicators to understand correlations used to forecast productivity and resourcing needs of a fast-growing business unit with little prior information. We also consider strategies to quickly integrate the new business into existing data platforms and storage, while managing multiple analytical upgrades within the whirlwind of new business expansion and on-boarding of new human resources.
Elizabeth Nielsen’s educational and practical experience is in applied mathematics. She currently creates business analytics for a new, fast-growing section of Quintiles: Real World & Late Phase Research. Previously, she led an analytical team that created models for the Operational Analytics group at Quintiles with a data-driven approach to create more efficient processes. Earlier in her career, she worked in industrial optimization and as an actuary in group health insurance. Elizabeth has an MS in Operations Research from the University of North Carolina, a BS in Mathematical Sciences from the University of Kentucky and is a Certified Analytics Professional (CAP).
In today’s fast-paced environment with consumer demands and product strategies changing based on a tweet or a snowstorm, additional flexibility and agility in the supply chain are more important than ever. We have developed an innovative approach to network optimization that can reinforce individual product strategies, achieve higher value, minimize risk, and better support business changes via more frequent evaluation. In this presentation, HGS will discuss their unique approach which combines network design, utilization, and active monitoring. By reviewing case studies in the food service industry, we will illustrate how this approach solves common challenges like:
- Network Optimization exercises done only once every 3-5 years can leave business exposed, unable to respond to ever-changing consumer demands, and can miss critical cost savings opportunities
- The dynamic nature of the business, where the solution can be outdated shortly after implementation
- The varying and different business strategies across product categories
Lisa Pell is a Senior Manager in the Analytics & Supply Chain Services division and has 20 years of supply chain and marketing analytics experience across multiple companies and industries. Lisa is currently leading the Network Innovation group at HAVI Global Solutions which focuses on bringing new optimization and analytic solutions to HGS clients. Lisa has a B.S. in Mathematics and an M.S. in Applied Statistics both from the Illinois Institute of Technology.
Nikhil Thaker is Senior Manager in the Analytics & Supply Chain Services division and has 10+ years of supply chain experience. Nikhil has been a consultant and worked with multiple companies and industry verticals throughout his career. Nikhil specializes in Optimization technology and has in depth knowledge of network, inventory and transportation optimization. Nikhil is CPIM certified with APICS and an active member of IIE (Institute of Industrial Engineers). Nikhil has a B.S. in industrial engineering from the Nagpur University, India, and an M.S. in industrial engineering from Oklahoma State University.
Optimization seeks to maximize or minimize an objective function under resource constraints. This is an elegant concept, yet when it comes to applying optimization in the “trenches”, it rarely comes out as that black and white. The problems present themselves without a hint of optimization, and yet, optimization turns out to be the ideal solution approach. For the past ten years, I have built many optimization models for my public sector clients. With the use of modern optimization modeling language, the elegance and flexibility of the modeling and the ease to communicate requirements are among many benefits to offer. This presentation will showcase some of my past optimization applications and thoughts on best practices with modeling and implementation.
Dr. Hua Ni, PMP®, CAP™ is a Senior Managing Consultant and Service Area Manager in the IBM Business Analytics and Optimization practice. Since joining IBM in 2004, Dr. Ni has developed a wide variety of advanced analytics and Operations Research solutions, including large-scale transportation optimization planning models, facility planning models, performance measurement system, and operational diagnostics models. His current focus is on deriving business insights and values through the analysis of large volume of data using advanced analytics techniques. Dr. Ni received a Doctor of Science degree in Operations Research from the George Washington University and was among the first group of awardees as a Certified Analytics Professional (CAP™). Dr. Ni currently serves as a trustee for the Washington DC INFORMS Chapter.
The negative impact of a disruption in a freight train network can cost millions of dollars. We develop a unique methodology using a multi-commodity network design problem structure to enable the handling of disruptions by optimizing the re-routing of freight trains. The model allows the user to consider the joint impact of re-routing several trains on the freight train network by quantifying the resulting effect on railcar delay. Our work includes applying the model to the real-world problem of solving disruption issues arising in the network of one of the four main railroad franchises in the United States.
Alborz Parcham-Kashani received his Bachelor of Applied Science in Industrial Engineering from University of Toronto, where he won the Ben Bernholtz Memorial Prize in Operations Research and graduated as a Shell Canada Scholar. He then moved to the Georgia Institute of Technology as a Wally George Fellow, where he received his Master of Science in Operations Research degree in May 2014. Alborz is currently working toward receiving his Ph.D. from the Georgia Institute of Technology in Industrial Engineering specializing in Supply Chain Engineering with a minor in Statistics. His research interests are in the area of applied operations research as it pertains to supply chain and logistics systems; specifically planning, design, and control in both deterministic and stochastic environments.
The United States Postal Service (USPS) delivers more mail to more addresses in a larger geographical area than any other post in the world. Consequently, in order to meet its service obligations, USPS must maintain a large network of contracted air transportation spanning 120 airports, over a dozen carriers, up to 12 million pounds a day, 7 days a week. In recent years, the United States Postal Service has leveraged a mixed integer linear program of their own design to provide service responsive comprehensive lowest cost transit solutions for this multi-billion dollar network. On a continual basis, rates and capacities of several carriers are balanced against the forecasted demand across a network of more than 9,000 different origin-destination pairs via a blending problem style linear program. This program continues to be a resounding success, providing significant savings and spurring subsequent optimization projects.
Brian Burns is a Network Operations Research Analyst within the Network Analytics department of the United States Postal Service. In his 6 years at USPS, Brian has worked on several strategic projects regarding the U.S. Postal Service’s mail processing infrastructure, retail equipment, and logistical networks. Brian earned his BS in General Engineering at the University of Illinois at Urbana-Champaign and is a certified Lean Six Sigma Black Belt. Brian currently supports air transportation purchase decisions and surface transportation network design.
9 - Analysis of Marine Corps Renewable Energy Planning to Meet Installation Energy Security Requirements
Our research analyzes Marine Corps energy consumption and renewable energy generation (REG) to develop REG metrics that support Department of Defense (DoD) energy security policies. Our research objective was to determine the costs of energy interruption and the net present value (NPV) of REG required to meet energy security objectives. To do this, we first evaluated the energy consumption and current REG practices of 14 installations. Second, we analyzed energy portfolios to calculate shortfalls from minimum energy requirements and then determined the cost to overcome shortfalls via REG technologies. Finally, we calculated energy interruption costs and determined the break-even point for REG options that meet energy security requirements. Through this study, we demonstrate that REG investment carries a positive NPV and increases energy security by maintaining baseline energy requirements critical for sustaining operations during an outage. While this study is Marine Corps specific, the methodology and lessons learned are broadly applicable to local, state, and federal government agencies as well as companies looking to minimize productivity losses from electrical grid interruption.
Captain Christopher Chisom has 7 years of active duty military experience. He is a Logistics Officer in the Marine Corps and has served in that capacity in a variety of Marine units. Currently, he is serving as a Defense Systems Analyst at Headquarters Marine Corps, Program Analysis and Evaluation (PA&E) division. His formal education includes an M.S. in Defense Systems Analysis from the Naval Postgraduate School and a B.A in International Policy from Miami University in Oxford Ohio.
10 - Decision Making under Uncertainty: Stochastic Model for Analysis of Biofuel and Food Production
Decision making under uncertainty is of crucial importance for obtaining long-term benefits in many investments. Uncertainties such as yield amount and price level in the market need to be considered for agriculture industry. In this study, we present a two-stage stochastic programming model to maximize revenue of food and biofuel production in the long run. This is the first model that incorporates uncertainties in decision making for competition between biofuel and food production. It also takes into account various economic and environmental effects of allocating different land types to food and energy crops. Application of the model and Benders decomposition algorithm is demonstrated by considering switching grass and corn production in a real case study in the state of Kansas. Our analysis provides significant insights such as profitability rates under various budget options for stake-holders in agriculture and biofuel industry.
Halil Cobuloglu is a research assistant and a PhD candidate in Industrial and Manufacturing Engineering at Wichita State University. He has hands-on experience in developing optimization, simulation, and neural networks models to derive business insights for biofuel production, renewable energy, and health care problems. He mentored three graduate students in their research projects. His articles are published in top journals in energy and decision science fields. Prior to Wichita State, he worked as a system analyst, project coordinator, and consultant for food, media, and business association. He earned a Master’s degree from Bogazici University and a Bachelor’s degree from Istanbul Technical University, majoring in Industrial Engineering. Mr. Cobuloglu is selected as the outstanding PhD level student for his success in research. He is also the recipient of the John F. Fargher Scholarship Award for having the best project and the second best oral presentation awards from WSU and IIE professionals.
From a long history of markdown sales, consumers have expectations that retailers will provide markdown discounts during a selling horizon. The consumer purchase timing decision is analyzed by using discounted expected utility theory, where consumers act to maximize their utility over time. The consumer’s sequential decision-making process is formalized under uncertain product availability. An optimal purchase timing policy is identified in a market environment, in which a strategic customer knows the markdown pricing scheme, available inventory level, and remaining time to the end of the selling horizon. For a given consumer product valuation distribution, the consumers’ purchase times are predicted by using the optimal purchase timing policy.
Ilhan Emre Ertan is a PhD candidate from The University of Texas at Dallas. He is currently working on his dissertation focuses on Strategic Consumer Behavior and its applications in the field of operations and marketing managements. He has an MBA degree and a BS degree in electrical engineering. Prior to his doctoral studies, he gained substantial work experience in several different industries including retailing, entertainment, consulting and engineering. He played critical roles in several projects when he was working for McKinsey&Co. and Blockbuster, Inc. He is very enthusiastic about playing in a blues band as well as being in a team to find innovative solutions to problems from industry.
This paper deals with customer churn prediction problem, which is typically modeled as a prediction problem of the next (feasible) time period. We propose a dynamic churn prediction framework for generating training data from customer records, and leverage it for predicting customer churn within multiple horizons using standard classifiers. The proposed framework includes customer observations from different time periods, and thus addresses the absolute rarity issue that is relevant for the most valuable customer segment of many companies. It also increases the sampling density in the training data and allows the models to generalize across behaviors in different time periods while incorporating the impact of the environmental drivers. As a result, this framework significantly increases the prediction accuracy across prediction horizons compared to the standard approach of one observation per customer; even when the standard approach is modified with oversampling to balance the data, or lags of customer behavior features are added as additional predictors. The proposed approach to dynamic churn prediction involves a set of independently trained horizon-specific binary classifiers that use the proposed dataset generation framework. In the absence of predictive dynamic churn models, we had to benchmark survival analysis which is used predominantly as a descriptive tool. The proposed method outperforms survival analysis in terms of predictive accuracy for all lead times, with a much lower variability. Further, unlike Cox regression, it provides horizon specific ranking of customers in terms of churn probability which allows allocation of retention efforts across customers and time periods.
Ozden Gur Ali
The County of San Diego Health and Human Services Agency provides services to approximately 1 million residents in our local facilities, including Family Resource Centers (FRCs). Predictive and trend analysis of our serviced population along with the Affordable Care Act revealed the need to evaluate our FRC locations. Utilizing Oracle Discoverer and Geographic Information System mapping software, we identified optimal locations for two new FRCs. As a continuing project, we are working on finding staffing ratios for each FRC by analyzing client data to fully understand the population we are serving. We will use this analysis to find the ideal worker needed at each location to better serve the community, taking into consideration language needs, travel time, and program knowledge. This analysis will also aid in determining square footage needed for services within the facility, and will be a model for future projects and expansions.
Jacqueline Hamed is an Operations Research Analyst and Jaime Mendez is an Administrative Analyst for the County of San Diego, Health and Human Services Agency. In their current positions, they use statistical and quantitative analysis, simulation and optimization techniques to analyze information and develop practical solutions to business problems. They are both Green Belt certified in the Lean Six Sigma process improvement techniques and have experience in project management. Jacqueline received her Master’s degree from San Diego State University and Jaime received his Bachelors from University of California, San Diego. They have over eight years of combined experience as data analysts.
Patients who fail to attend appointments (no-shows) and patients who cancel their appointments in advance (cancellations) may complicate appointment scheduling, and may cause disruptions to an outpatient healthcare clinic. We employ individualized models for predicting patient no-shows and patient cancellations based on patient demographics and past patient behavior, and use those models to inform the scheduling process. The scheduling model is still in development; we are currently analyzing how the original no-show and cancellations models should be altered, or variables that should be included, to make the models more suitable for a scheduling application.
Shannon Harris graduated with a degree in Systems Engineering and Operations Research from George Mason University in 2007. She worked at Deloitte Consulting from 2007-2009, and as a Cost Analyst at Technomics, Inc. from 2009-2011. In 2011 Shannon began pursuing a doctoral degree in Business Analytics and Operations at The University of Pittsburgh Katz Graduate School of Business; her anticipated graduation date is August 2016. Shannon’s research involves developing predictive models using data mining and statistical techniques in order to improve healthcare operations and best practices.
Recently, as many business processes are automated, companies can obtain huge event logs accumulated in a variety of business information systems. Such event logs can be utilized to keep track of conducted tasks and context in business process execution. By analyzing the event log, we can investigate and understand business activities better, considering their underlying business processes.
In this research, we propose a framework for operational analytics by using business process mining techniques. Specifically, the detailed procedure of business process analytics is presented based on four phases of analytics: descriptive analytics, diagnostic analytics, predictive analytics, and prescriptive analytics. A concrete example is also provided by analyzing real-life event log in a service industry. The proposed framework can be adopted to support precise diagnosis, forecasting and treatment of business problems.
JinSung Lee received a BS and an MS from the Department of Industrial and Management Systems Engineering at Kyung Hee University, Yongin, Republic of Korea in 2007 and 2010, respectively. He is currently a PhD candidate in the same department. His research interests include business process management, process mining, business analytics, and service science.
Given the drastically changing healthcare landscape in the US today driven by health reform regulations and technology growth, particularly that of social media, health plans are exploring new and creative methods to reach out to their members to offer health information, provide support and encourage healthy behaviors. This research project involves a case study based on one of the largest health plans in Pennsylvania that explored the propensity of its members to adopt various types of social media platforms for their health and wellness management. Data were collected via an online survey on a host of key factors such as demographic factors, clinical factors (including the presence of a chronic condition or not), technology-based factors (like the frequency of social media and online activities engaged in), factors related to privacy and security concerns, etc. The insights gained from this study could potentially help health plans design and deliver cost-effective healthcare solutions to targeted segments of the population.
Dr. Sinjini Mitra is an Assistant Professor at the Department of Information Systems and Decision Sciences in the Mihaylo College of Business and Economics at California State University, Fullerton. Prior to this, she was a Postdoctoral Research Associate at the University of Southern California’s Information Sciences Institute. Dr. Mitra received her Ph.D. degree in Statistics from Carnegie Mellon University. Her research interests include data mining, business analytics and statistical applications in healthcare, education security and information systems. She has more than 30 research publications, including 11 peer-reviewed journal articles and 10 book chapters. She teaches courses in the area of Statistics and Analytics.
In many code schemes for manufacturing parts or raw materials, the item codes reflect the specifications of items, such as the size, quality of material, intensity, and color. In materials requirement planning, text retrieval, such as searching for compatible items, substitute items, or successor items for discontinued lines, is necessary under the item code schemes described above. A company with multiple manufacturing suppliers must deal with different code schemes during text retrieval. This makes it difficult to execute versatile text retrieval algorithms to perform searches within a reasonable amount of time and obtain highly accurate results. We modified the existing edit distance and metric tree search rules such that it can be applied as a similarity measure for more than two million items based on multiple code schemes; this reduces the retrieval time to 1/10 of that of the previous retrieval system.
Kohji Molikawa is a researcher at the Research and Development Department at Hitachi Solutions East Japan, Ltd. He received his BSc, MSc, and PhD in Astrophysics from Tohoku University, Japan. He is a certified Software Design and Development Engineer at Information Technology Promotion Agency, Japan. He has more than ten years of experience in analytics applications such as price optimization at retail, text mining, and approximate string matching. He is a member of the Institute of Electronics, Information and Communication Engineers, Japan.
This work is concerned with the development of a response model in the aftermath of a Mass-Casualty Incident (MCI). It presents various OR techniques and predictive analytics for solving realistic MCI scenarios. A rigorous mixed integer programming (MIP) formulation is proposed for solving the combined ambulance dispatching, patient-to-hospital assignment, and treatment ordering problem. Both simulation and optimization based approaches are presented; including MIP-based construction heuristic and iterative local search metaheuristic algorithms. The objective is to minimize the time required to provide emergency treatment to all patients. The proposed model is challenged on the hypothetical case of a terror attack at the New York Stock Exchange in the Lower Manhattan with up to 150 trauma patients. The bottlenecks for various capacity settings are identified, in terms of the number of ambulances and available hospital beds, while the effect of including remote hospitals as opposed to reduced ambulance transportation times is illustrated.
Dr. Repoussis is Assistant Professor of Operations at the Howe School of Technology Management (Stevens Institute of Technology). He holds a Diploma in Chemical Engineering (National Technical University of Athens), a MSc in Process Systems Engineering (Imperial College) and a PhD in Management Science and Technology (Athens University of Economics & Business). His research focuses on the design and development of mathematical models and algorithms to aid decision making in operational planning and scheduling of resources. He has been involved with the organization of several conferences, and he has served as member of the BoD for the Hellenic Operational Research Society. He has been also involved in various NSF and EU FP7 projects, and his research has been funded by non-profit organizations and private companies.
The client was evaluating new policies and strategies to shift the petroleum portion of US transportation energy consumption from 93% to 50% before 2040.
PwC developed an integrated model combining TCO, consumer adoption, petroleum consumption, and economic analysis.
PwC enabled the client to capture and quantify the feedback between consumer adoption, manufacturer production, marketing, and the development of refueling infrastructure.
PwC simulated the outcomes of proposed strategies across a range of uncertain future scenarios, recommended a set of KPIs to track, and the combination of policies most likely to be effective as the cost outlook for each technology platform crystallizes.
Ryan Hughes is a senior associate in PwC’s Advisory service, has an MBA/MSc degree in System Dynamics, has over five years of consulting experience, and has presented at INFORMS twice.
The increasing demand for air travel over the years has put United States airports to their capacity limit that results in risk of flight delays which lead to negative impacts on the aviation industry, airlines reputation, and airport operations. Understanding and mitigating flight delays in the National Airspace System (NAS) is a major long-term objective of the Federal Aviation Administration (FAA). The complexity and dynamism of NAS creates great difficulties for airports in handling the flight delay incidents and these challenges have not been addressed in the existing literature. The purpose of this study is to develop a more robust prediction model of flight delays in such system. Results of this study can support airport authorities in making decisions to mitigate the flight delay incidents.
Dr. Dothang Truong is an Associate Professor of Doctoral Studies at Embry Riddle Aeronautical University. He is Certified Supply Chain Professional (CSCP) by Association for Operations Management (APICS). He teaches data mining, operations research, advanced statistics, and supply chain management. His research interests include predictive modeling, sustainable supply chain management, cloud computing, and e-procurement. He has established a strong record of publication in top-tier journals including International Journal of Business Analytics, International Journal of Electronic Market, Journal of Business-to-Business Marketing, Journal of Organizational Computing and Electronic Commerce, Industrial Marketing Management, International of Entrepreneurship, Journal of Enterprise Information Management, among others.
21 - A Perturbative Clustering Hyperheuristic Framework as Region Splitter for Danish Railway System
New signaling system in Denmark should ensure fast and reliable train operation, having very strict time limits on recovery operations. Moreover, the new signaling system is based on completely different hardware, requiring different maintenance tasks. This makes it interesting and necessary to rethink whole preventive and corrective maintenance scheduling. Due to geographic features of Denmark, crew is physically located in different locations than being settled in one depot. This calls for a more precised assignment of maintenance tasks to the crew in terms of distance to avoid a bad planning (high total distance cost) or even an impossible one. We suggest that the biggest maintenance area in Denmark could be partitioned into subregions prior to scheduling phase, considering the tasks and crew locations. To minimize overall distance cost in preventive maintenance planning and maximum availability time in future breakdowns, we propose a perturbative clustering hyperheuristic framework as region splitter. The framework introduces ve low-level heuristics and employs an adaptive choice function as a robust learning mechanism to lead the framework towards a better search space. The framework improves a complete solution by reassigning far away tasks (outliers) to a better cluster choice at each iteration while taking balanced crew workload into account. The behavior of the proposed low-level heuristics in the adoptive choice-function based hyperheuristic is investigated on 12 dataset and results are compared with random hyperheuristic. The adaptive clustering hyper heuristic framework could obtain roughly 13% improvement at both total driving cost and maximum availability on a middle quality solution and 10% of better results compared to random clustering hyperheuristic. Finally, to assess cohesion of the clustering results to be used in scheduling phase, validity factor of compactness was measured which resulted in 32% improvement by presented framework.
Shahrzad Mohammadpour is a young researcher working on hyperheuristics and data clustering research areas for solving real world scheduling problems. Since 2013, she started her PhD on railway maintenance planning at the Technical University of Denmark. Shahrzad received her Bachelor’s of Software Engineering at Buali Sina University, Iran in 2007 and a Master’s of research in computer science at the University Putra Malaysia in 2010. Her academic background helped her to be very familiar with Local Search Based Techniques and meta-heuristics – particularly, hyper-heuristics and timetabling problems. After earning her Master’s degree, she worked as a software developer in a banking specialist company in Kuala Lumpur, Malaysia for two years. The achievement of her working experience helped her to be professional in object oriented programming and software development fields. She loves the research areas of computer science which aim to fill the gap between theory and practice.
Business schools are under growing pressure to engage in significant programmatic reforms in light of the business community’s call for web-savvy, problem solving graduates. Even AACSB has gotten into the reformation act by recommending the adoption of a comprehensive collaboration strategy. To meet these and related financial challenges, many schools of business are turning to Analytics. Typically in a graduate management setting upwards of two-thirds of incoming students do not have a business or scientific undergraduate background. Analytics based learning systems can be used to assist in curriculum design and student intervention initiatives through intelligent tutors. Typically, the learning goal should be to develop general analytic and quantitative problem-solving skills. Some specific benefits for adopting Analytics throughout schools of business include: 1) enhancing student recruiting/retention; 2) addressing the growing demand for Analytics talent; and 3) improving learning outcomes. The purpose of this poster presentation is to highlight best practices in the use of Analytics throughout the management education universe.
Dr. Owen P. Hall, Jr. holds the Julian Virtue Professorship and is a Rothschild Applied Research Fellow. He is a Full Professor of Decision Sciences at the George L. Graziadio School of Business and Management, Pepperdine University. He is the recipient of the Charles Luckman, Howard White, and Sloan teaching excellence award. Dr. Hall has over 35 years of academic and industry experience. He is a registered professional engineer, State of California. Dr. Hall has written extensively on cloud based collaboration, data mining and hybrid learning. He recently received a grant from GMAC to design a collaboration network for enhancing management education. Dr. Hall received his Ph.D. from the University of Southern California and undertook post-doctoral studies at the Center for Futures Research.