Session Abstracts(listed alphabetically by speaker last name)
Predictive Analytics and Forecasting Patient Demand: Saving Lives in the Emergency DepartmentPatient-care needs in the Emergency Department (ED) are highly variable, often deviating significantly from "average" at any given moment. Reacting to this rapidly changing environment with staff schedule adjustments is only marginally successful, if helpful at all. Frequent mismatches between patient need and the healthcare professionals available to provide that care can result in significant human and financial costs.
Using SAS/ETS software, we have developed a time series framework that combines multiple cycle effects (e.g., hourly, day of week, monthly) with a baseline component to produce ED demand forecasts at a one-year horizon. This approach accommodates cycle collinearity and techniques for handling nonstationarity and feature (lag) selection in a way that emphasizes extrapolation performance. The result is a robust approach that enables the planning of health care professional resources to meet highly variable patient demand. Using key patient attributes we are able to predict the nurse and physician workload needed to meet the future needs of patients in a given ED.
Looking Beyond Traditional Data Sources to Improve the Customer ExperienceThe Marketing Analytics department at OgilvyOne is constantly looking for innovative ways to augment existing client data in order to improve and enrich the customer experience.
Historically, OgilvyOne looked to traditional channels for adding new data, using third-party data providers to append demographic and lifestyle data. While these channels are effective for providing the data needed for modeling and data mining projects, they aren't always cost-effective.
As a result, it recently started using free data sources and APIs that provide an even richer set of data. Although much of this data is unstructured and comes from many different sources, it can be used to greatly improve customer engagement. For example, OgilvyOne now uses census, Twitter and Facebook data to help analyze its clients and target markets.
This presentation will highlight OgilvyOne's use of free data feeds to create a highly targeted weather-based communication to help its clients sell more seasonal products. In a recent project, OgilvyOne analyzed historical data for US weather in order to create market segments based upon target temperatures for each segment. Then, the segmentation was tested and resulted in 131 percent over 2011's performance.
Making Strategic Decisions About the Role of Business Analytics in Your CompanyData analytics and big data are quickly becoming the next wave in management improvement strategies. The momentum is evident, but the impetus and business justifications, though important, are vague and uninspiring. Executive leadership, strategic integration and application, and alignment with organizational strategies are vital to successfully leveraging data analytics and capitalizing on the opportunities big data present. As such, in this talk we will focus on trends in data analytics. We will also discuss various levels of strategic involvement, with respect to data analytics, and their potential impact. Additionally, we will provide guidance on how to determine the appropriate level of strategic involvement for your organization and how to assess your readiness to achieve that level.
Adding Power to Optimization with SAS/OR and High-Performance ComputingSAS/OR provides a powerful and versatile set of optimization resources, and high-performance computing opens up even more possibilities in optimization. PROC OPTMODEL enhances its nonlinear optimization abilities by adding a high-performance multistart optimization algorithm, designed to find better solutions for nonconvex problems. PROC OPTMODEL also adds a high-performance decomposition algorithm that exploits a frequently occurring structure in linear and mixed integer optimization problems to accelerate the solution process. Finally, a new local search procedure uses high-performance capabilities to implement a parallel heuristic approach with optimization problems that are difficult or impossible to solve with classical optimization methods. We'll discuss these and other advances that increase the power and scope of optimization with SAS.
Revolutionizing Decision Making: How Analytics Will Take Over the BusinessWith advances in big data, artificial intelligence and increased metric captures of everything we do, analytics will go through a radical transformation in the next few decades. As a result, there will be a shift from analytics simply influencing business decision makers to analytics actually owning the decisions. This transformation is already happening in the pricing community, but expect it to expand to even CFO-level decisions.
Consumer Trends and AnalyticsConsumers are continuing to place greater emphasis on the consumption "experience" of goods and services. For instance, vacationers are increasingly attracted to customizing experiences that are tailored to their individual needs and tastes. Many companies strive to empower consumers by developing initiatives, tactics and tools to optimize offerings and assist consumers in creating customized experiences. In addition, vacationers are also looking to customize the end-to-end experience, which is in contrast to customizing a collection of episodic individual experiences. We present an interpretation of this trend and provide a perspective on implications for analytics and decision sciences.
If it Looks, Swims and Quacks like a Duck...; Using Predictive Analytics to Target Taxpayer Fraud and ErrorRevenue, the Irish Tax and Customs Authority, has been using the power of data mining techniques as part of a process of putting analytics at the core of its business processes. Predictive models currently in production include those targeting noncompliance/tax evasion, likelihood to yield, and liquidation. Currently Revenue is employing its Real Time Risk Framework, which scores cases for their riskiness in the live transactional systems, for both PAYE (Pay As You Earn) and VAT (sales tax) taxpayers. These initiatives have led to real savings for the exchequer, and the use of analytics in Revenue is one of ongoing consolidation and improvement. This paper describes Revenue's experiences using analytics, highlighting successes, lessons learned and possible future developments.
Is Winning a National Championship a Good Predictor for Undergraduate Enrollment?The "curse" of a championship spreads all over campus and the nation, but does winning all those championships really affect the student recruitment efforts and therefore the size of a freshmen class? The University of Alabama disputes this myth with a historical timeline of events that have shaped the class more than games on the gridiron. Discussion topics will include tools used to track student interactions, state funding issues versus a plan for recovery, and a business analytics approach to examining enrollment figures versus athletic team records.
The Next 'Big' Thing in Data MiningWith all the buzz about big data, and in response to customers' needs to work with larger and larger data sets, considerable advances have been made over the last few years to take new and existing modeling methods from statistics and machine learning and take advantage of more modern threaded and distributed architectures. With the release of SAS Enterprise Miner 12.3 on SAS 9.4, this new functionality is available through new nodes for sampling, exploration, modification and modeling. The use of the procedures is also enabled directly. Join us as we discuss the additional methods, new features, best practices and future development plans.
Text Analytics and Latent Semantic DimensionalityThe emergence of big data analytics has generated a lot of interest in the quantitative analysis of unstructured text data. Customer comments, news stories, industry report segments, tweets and email messages are now routinely analyzed by text mining software solutions such as SAS Text Miner. Latent semantic analysis, a text analytic framework for extracting conceptual dimensions, offers solutions for analytic needs, including include document summarization and incorporation of unstructured text into quantitative predictive modeling. This presentation addresses the problem of latent semantic dimensionality selection. From simple visual examination of eigenvalue scree plots, to the implementation of an algorithm for multiple elbow point detection, the presentation will cover the detection of multiple dimensionalities in textual data. A number of illustration examples will show how document collections - including responses to open-ended surveys, customer comments and industry report paragraphs - can be examined at alternative levels of semantic abstraction that represent topics, megatopics and microtopics.
Integrating SAS with Optimization ModulesFor a specialized project, it may be necessary to integrate the functionality of multiple analytical tools with differing areas of strength. In this presentation, we will discuss how the analytical capabilities of SAS can be combined with the capability of an external module, such as a custom heuristic for a hard optimization problem.
Solving Price Optimization Models with SAS/ORThe numerical solution of price optimization models can be complicated by factors such as a nonlinear demand curve or dependencies between multiple products. We will present results from testing SAS/OR on selected price optimization models from published research literature.
Forecast Value Added: A Reality Check on Forecasting PracticesMany organizational forecasting practices have little positive effect, or may even be making the forecast worse but how would we ever know? This presentation shows how the Forecast Value Added (FVA) metric is being used by many organizations to identify non-value adding steps and participants in a forecasting process. Focusing on the overall effectiveness of the process, we will see how to
- set appropriate expectations for forecast accuracy
- gather the data necessary for conducting FVA analysis
- interpret FVA results and communicate results to management
- identify and eliminate forecasting process waste.
Finding Members in Need: Algorithms Using SAS, Rules Engines and Predictive Modeling to Identify Care Management OpportunitiesHealth plans continue to work toward finding more effective ways to identify opportunities to deliver value to their membership and remain relevant in the health care arena. While sharing EHR data between health plans and payers is still a distant hope, there are opportunities now to do more with administrative claims data through rules engines and predictive modeling. Through analytics and outcomes analysis, CareSource has developed a model that codifies key indicators and can identify a member in need of complex case management. Using SAS Analytics, various rules engines can be incorporated into the model, which creates a series of "valves" that can be fine-tuned over time in order to pinpoint the right mix of members who require more intensive care management. This process supplements nursing interactions by incorporating claims history, case management notes and other relevant information. This session will focus on the purpose of the model for CareSource, how it was built and the various valves that allow the model to be flexible and adaptable.
Credit Loss Forecasting and Stress Testing using Transition MatricesAs financial institutions are recovering from the recent recession, Credit risk continues to be a major component of the overall risk for commercial banks. In the context of the new regulatory requirements they are faced with, banks need models not only to forecast losses, but also to evaluate their impact on capital. Forecasting methodologies using credit-transition matrices provide unique advantages in this context but come with their own challenges. The Capgemini team will share some perspectives on this topic based on their client project experiences.
Actionable Big Data Analytics for Health Care and Life SciencesWith the steep rise in health care costs and changes in policy and regulation that affect financial reimbursement and patient-care performance evaluation, organizations are forced to move quickly to improve data analysis. Electronic health records, emerging health information exchanges, and other data services are providing new and valuable data. Highly visual and easily understood data presentation and analysis are now vital to many financial, clinical, and patient-care decisions.
Which of my patients are at risk for poor outcomes? When, and why, does my health care organization provide suboptimal care? How can I allocate resources to improve patient care and lower costs? Every member of the global health care system is actively engaged in answering these questions, from the individual clinician to the care manager to the chief executive of the largest integrated delivery network.
In this presentation, we will share our knowledge from recent experiences by taking a look at the current analytics environment in health care and demonstrate why big data is the way forward. We will then clarify why it is critical to tightly integrate data center technology that is optimized for scalability, security and performance with a clinically intelligent platform that can make sense of a huge variety of data types from multiple sources.
Developing an Analytic Road Map Do's and Don'tsCox Communications is the third-largest cable provider in the US, with nearly 6 million customers, and is noted as an industry leader for its high-capacity, reliable broadband delivery network, its integration of Internet and television into a single compelling product, and its superior customer care.
As one can imagine, Cox has access to an incredible variety of big data sources, and the potential to apply these data sources across a large number of applications. Nearly all of these data sources are potentially valuable - but in a world of limited time and resources, the challenge is in determining how to prioritize efforts and then deliver analytic work products in a structured, prioritized way to deliver the greatest amount of value. In other words, our challenge isn't in identifying where analytics can help us, but rather in ensuring that we're always working on the "right" problems where we have the most leverage so that we can deliver the most value to the business. Using specific examples and case studies, this presentation will demonstrate our structured process for prioritizing our efforts to create our analytic road map at Cox - both in terms of what we are taking on and what we decided not to take on - and some outcomes of our initial efforts.
Analytics-Empowered Pharmacy Benefit Managers (PBMs) to Improve Member EngagementPharmacy benefit managers (PBMs) have for years relied on competitive analysis, a focus on drug pipelines and negotiated rebates as primary strategies to optimize the performance of their pharmacy segments. Despite best efforts, industry analysts believe the 2011 downturn in mail pharmacy prescriptions is part of a multiyear trend. The slowdown is also reflected in the mail penetration rates at the big three PBMs. Appropriately, given that outlook, PBM executives are beginning to take a more comprehensive look at the factors influencing pharmacy performance and seeking tools to more effectively improve their decision support on the way to better health outcomes and streamlined costs for patient care.
CVS, the largest pharmacy health care provider in the United States, with integrated offerings across the entire spectrum of pharmacy care, has uniquely positioned itself in this market downturn. CVS has partnered with Accenture's advanced analytics practice to engage plan members in behaviors that improve their health and to lower overall health care costs for health plans, plan sponsors and their members. Among many sophisticated analytics techniques used to improve member engagement is a joint experiment to infuse members' mobile usage data into predictions of members' propensities to respond to marketing programs and habitual adherence to drug treatments.
A 'Hippocratic' Model for Analytics in Health Care"Declare the past, diagnose the present, foretell the future." Hippocrates Constant innovation and change are hallmarks of health care. The changing paradigms of regulation, payment mechanisms, consumerism and competitive pressures are forcing providers to adjust to new priorities, and redefine themselves using new standards. Successful organizations are differentiating themselves through comprehensive management strategies, transforming their culture to value and use data and analytics, investing in infrastructure, and digesting and interpreting new information channels so they can be used to confidently make decisions. The Cleveland Clinic employs a comprehensive analytics strategy that emphasizes resource deployment mechanisms, meaningful content, data governance, and user accessibility and capability. Case studies will highlight this strategy and its evolution to achieve Hippocrates' 5th century challenge.
Turning Big Data into Localized Intelligence
One promise of big data is to provide the required scale to offer up localized insight on dynamic events - the proverbial data fire hose with sufficient flow for faster tracking of product adoption or identification of competitive threats. In this session, learn about tools and techniques being used to turn big data into localized intelligence. Topics include measurement of data velocity, clustering to discover new patterns and market-share trending.
Speech Analytics Applications to Predictive ModelingSpeech or voice analytics is an emerging technology that is gaining the interest of contact center operators, as it provides insights into a previously untapped, yet massive, source of information, which is the true "voice of the customer," i.e., the recorded phone calls. The potential range of applications of such technology is impressive: QA automation, 100 percent compliance adherence, call center metric (CSAT, AHT, Conversion, etc.) optimization through driver discovery, desktop performance optimization through targeted agent coaching, call model assessment, and predictive modeling. In this paper, we focus on the speech analytics technology application to improve the lift of propensity-to-pay models in an industrial setting by incorporating dynamic payment behavior indicators into existing static models. This approach is similar to using other types of unstructured data, yet quite different, in the sense that no text mining is performed on call recording transcriptions. The indicators are derived by the speech analytics tool by analyzing combinations of phrases and behaviors without the use of transcription. This proposed solution optimizes call center costs while concentrating agents' efforts on the most lucrative accounts and addresses the challenge of static treatment plans by listening to customer responses. Results pertaining to expected lift are provided.
Predictive Analytics in Social Media and Online Display AdvertisingThe last decade has seen unprecedented growth in the space of online advertising and digital media marketing. The new wave of social media (Facebook, Twitter, etc.) is making it easier than ever for marketers to reach the right customers at the right time with the right products and offers. However, the marketers, online advertising platforms and other stakeholders need to be equipped with suitable analytical tools and methodologies to maximize the potential of online and digital media. The traditional analytical tools are often insufficient due to the rapidly growing volumes of data as well as the increasing importance of dealing with textual and unstructured data. In this talk, we will present three case studies on applying data analytics in social media and online display advertising to help our clients stay competitive in the marketplace.
Parallel Processing in SASAs the cost of purchasing computing power (i.e., processors and RAM) goes down, there is a growing opportunity to gain access to high-performance servers or clusters. These environments make parallel processing on a larger scale a real possibility. There are two primary challenges in implementing a multithreaded approach from a practical standpoint: 1) identifying the parallelizable components of the program and 2) coding these components. This talk provides a brief overview of both of these topics and highlights the relatively easy-to-implement parallel processing functionality provided by SAS/CONNECT.
Speeding the Analytical Life Cycle: A Statistician's Perspective on Solving Problems with SAS Visual AnalyticsThere are high stakes on statisticians' ability to quickly build mathematical models that help solve business problems. Yet the early stages of an analytical project can consume most of the project's allotted time - leaving little time for model building. Why?
The answer relates to the uncertain and iterative nature of many tasks in the analytical life cycle. For example:
- Many data sources have unknown or hidden value.
- The available data often does not support the analysis.
- The data may be too large for statisticians to explore patterns holistically.
- Data extraction requirements are often based on intuition, not facts.
This presentation will present some ideas on how SAS comes into play with SAS Visual Analytics. We will discuss how you can formulate your business problem correctly and comprehensively, and then identify the relevant data for building good analytic models. And we will show you how you can do it all faster than ever before, with potentially great results for your business.
We Are All BayesiansDuring the last two decades the use of Bayesian methods in marketing analytics has grown significantly. Bayesian analysis is being used to solve a wide range of marketing problems, including product-launch forecasting, price optimization and media mix optimization. While the conceptual framework of Bayesian modeling has long been recognized, the recent growth in popularity can mainly be attributed to breakthroughs in computational techniques and modeling methods that make Bayesian methods attractive for large-scale, real-life modeling problems.
In this talk we will focus on Bayesian regression - i.e., regression models where the coefficients are assumed to follow random distributions that can be influenced by "outside information." Despite the growth in the popularity of Bayesian methods, there are still detractors who denounce Bayesian regression and claim that such models are fabricating results and not being true to the data.
However, this is not true. On the contrary, it turns out that Bayesian regression is an extremely powerful technique that can overcome issues that cannot be resolved with classical modeling. In fact, I will submit that once you discover the power of Bayesian regression, you will realize that you were a Bayesian all along.
In addition to showing a real-life case study, we will discuss some compelling benefits of Bayesian regression:
- Combating multicollinearity
- Reducing the chance of over-fitting
- Building granular models that actually make sense.
Using SAS Visual Analytics to Forecast Patient Visits at Hundreds of Medical PracticesCarolinas HealthCare System delivers care in nearly 800 separate locations. To ensure that patients have the proper access to care and that clinical resources are efficiently utilized, CHS forecasts patient volumes at each location daily for the remainder of the current month. Time series models produce accurate forecasts at the practice level, but no single model performs well across all practices. Developing individual practice-level time series models and deploying the model scoring logic in existing reporting applications could take one statistician a full year. SAS Visual Analytics allows CHS to automate model selection at the practice level, and present the forecasts and associated statistical control charts to practice staff and corporate management in a rapid and visually compelling manner.
Content Categorization a Road MapBuying a product to classify corporate content is just the start of the process. Given enterprise endeavors can have their own complexity, an organization implementing the SAS Enterprise Content Categorization application might not know where to start. Based on real-world examples, this session aims to provide practical guidance on implementing categorization for high-volume document classification. The session will include analyzing the goals to be accomplished with the software, how to define and then measure success, and strategies for managing the implementation. This session is geared toward those with little or no prior experience with the SAS categorization technology.
Improving Electricity Forecasting in a Volatile Market - The Journey of RWE npowerWith UK energy demand becoming ever more volatile and the costs of forecast error increasing, RWE npower faced the challenge of improving its time series prediction capabilities. Our presentation takes the audience through the journey RWE npower has embarked on in accurately forecasting UK energy demand and fine-tuning statistical forecasting algorithms for electricity, gas and wind demand, as well as processes and systems. To go beyond conventional approaches, RWE npower has been working collaboratively with the Lancaster Centre for Forecasting to evaluate enhanced forecasting models of artificial neural networks, in order to deliver further value through nonlinear energy demand forecasting. To integrate it all, RWE npower has implemented a new forecasting platform run by SAS, including model management across life cycles, that provides greater flexibility and responsiveness to changing market conditions.
In this talk, you will learn about the approach taken, technical capabilities implemented and benefits gained with the new forecasting platform.
Utility Customer Payment Arrangements: Analyzing Past Performance and Predicting Future SuccessWhen customers get behind on bill payments, they have the option of making special payment arrangements to avoid having their electric service interrupted. This presentation will 1) give insight into payment arrangements with residential customers that were made over a 12-month period, assessing their success across volume, amounts and duration, 2) evaluate policy changes that could reduce the number of payment arrangements and associated arrears and 3) show how a behavioral model can be developed to predict the outcome of a customer's next payment arrangement.
Time Series Data Mining: A Retail Application Using SAS Enterprise MinerModern technologies have allowed for the amassment of data at a rate never encountered before. Organizations are now able to routinely collect and process massive volumes of data. A plethora of regularly collected information can be ordered using an appropriate time interval. The data would thus be developed into a time series. With such data, analytical techniques can be employed to collect information pertaining to historical trends and seasonality. Time series data mining methodology allows users to identify commonalities between sets of time-ordered data. This technique is supported by a variety of algorithms, notably dynamic time warping (DTW). This mathematical technique supports the identification of similarities between numerous time series. The following research aims to provide a practical application of this methodology using SAS Enterprise Miner, an industry-leading software platform for business analytics. Due to the prevalence of time series data in retail settings, a realistic product sales transaction data set was analyzed. This information was provided by dunnhumbyUSA. Interpretations were drawn from output that was generated using "TS nodes" in SAS Enterprise Miner.
Non-Empirical Modeling: Incorporating Expert Judgment as a Model InputIn business environments, a common obstacle to effective data-informed decision making occurs when key stakeholders are reluctant to embrace statistically derived predicted values or forecasts. If concerns regarding model inputs, underlying assumptions and limitations are not addressed, decision makers may choose to "trust their gut" and reject the insight offered by a statistical model. This presentation will explore methods to convert potential critics into partners by proactively involving them in the modeling process, and incorporating "simple" inputs derived from expert judgment, focus groups, market research or other directional qualitative sources. Techniques include biasing historical data, what-if scenario testing, and Monte Carlo simulations.
Looking at a New Look: SAS Contextual Analysis and SAS Visual Analytics Help You Understand Your Unstructured DataSAS Contextual Analysis and SAS Visual Analytics are powerful companions in helping derive value from unstructured data. Understanding the knowledge trapped inside unstructured data can be a daunting task. SAS Contextual Analysis provides the tools to extract, explore, understand, model and capture this knowledge. Once captured, the powerful visualization tools provided in SAS Visual Analytics enable quick and easy data exploration, delivering valuable insight and useful information.
The Future of Digital Targeting and Measurement in a Big Data WorldThere has been an explosion of new digital advertising media in the last 10-15 years. At the same time, there has been an equal amount of growth in consumer data being collected. Advertisers and marketers have been slow to take advantage of the data to help inform targeting and measurement of these new media opportunities. Historical measures of success like click-through rates and impressions, were fine when that was all that was available. However, in today's world of granular big data, you can measure advertising with observed behavior changes as the measure of success. Individual- or household-level data provides the ability to dissect results across different customer groups to better understand who was really affected by the advertising. It can also be used to create one-to-one targeting in areas never before possible. Our research in the consumer packaged goods industry across mobile, online, social, TV, and multichannel media illustrates a new level of analysis that can be accomplished now that the technology in the media space and the consumer data capture can be merged into an unprecedented single-source data set.
Marketing Mix Models Not just for Media Mix AnymoreAT&T consistently invests over $1 Billion in advertising per year, to consistently rank among the top advertisers in the U.S. To manage this investment, AT&T has deployed market mix modeling to optimize media mix, with NPV impacts of these efforts exceeding $200M per year.
However, market mix modeling can drive insights and influence decisions beyond media mix. AT&T has applied market mix modeling techniques to influence other types of decisions.
This session will explain the foundation of AT&T's market mix modeling approach. It will then expand beyond media mix to show how the same tools provide guidance on drivers of customer churn, efficiency of TV advertising by network and management of mix of advertising messages.
Graph Optimization and Human Mobility AnalysisAs a highly competitive marketplace, the telecommunications industry is continuously evolving. Understanding customer behavior is crucial to better compete in such an environment. Social network analysis can play an important role in creating a complete picture about customers' behavior, not just in terms of their individual patterns but also in relation to their relationships throughout the network. The analysis of subscriber mobility also reveals customer behavior in terms of paths and trajectories mostly traveled, as well as how they relate to other subscribers not just in terms of frequency, but also in respect to the geographical locations. Are the communities geographically close? Are customers who are frequently connected close or distant from each other? In addition to understanding customer behavior, the human mobility study may increase understanding of areas like traffic optimization, network planning and the spread of disease.
Education Panel SessionThe panel includes:
- Michael Rappa, Executive Director of the Institute for Advanced Analytics and Distinguished University Professor of Computer Science, North Carolina State University
- Jeff Camm, Chair of the Department of Operations, Business Analytics and Information Systems and Professor of Quantitative Analysis, University of Cincinnati
- Goutam Chakraborty, Professor of Marketing, Oklahoma State University
- Dr. Charles Sox, ISM Department Chair, University of Alabama
- Richard Sowers, Professor of Mathematics, University of Illinois at Urbana-Champaign.
Big Data, Academia and the Job Market: How Universities Are Minding the GapWe have all read the headlines and heard the statistics - there is a gap between the demands for analytical talent and the supply - and it's getting bigger. So what are universities doing about the gap? The answer is a lot - but maybe not in the ways that people might think.
This panel session, which includes some of the country's top faculty in the area of applied analytics, will focus on how universities are minding the gap.
Do You Know or Do You Think You Know? Creating a Testing Culture at State FarmExperimental design is becoming more common in business settings, both in strategic and tactical arenas. Firms are beginning to recognize the broad applications, including initial product design, marketing, logistics, customer retention, profitability analysis, strategy optimization and website design, just to name a few. Surprisingly, many still resist fully integrating designed experiments into their strategies.
As with any analytic change, driving an organization to adopt strategic testing can be difficult. The presenter will share what he's been doing at State Farm to move the firm from minimal business experimentation to a more ambitious culture of testing, as well as provide a step-by-step guide for changing or creating a testing culture. He will share the simple example from State Farm's online presence that demonstrated the value of multivariate testing to the firm and also the ambitious comprehensive optimization test this success launched as a follow-up, as well as the current state of the constantly evolving testing landscape.
The presentation will provide some tangible takeaways for business practitioners in the crowd desiring to improve their own strategies through designed experiments and is expected to be beneficial to both experienced experimenters and those who want to champion testing in their companies but are unsure how to get started.
Analytically Building Your Analytics Bench: 2012 Analytics Professionals Study ResultsMany innovative businesses and IT organizations appreciate the competitive advantage analytics capabilities can provide and have ambitions to reach increasing levels of analytics maturity. However, the well-documented shortage of analytic talent leaves many firms without a strong analytic talent bench and little knowledge about how and where to find analytics professionals needed to get there.
In this presentation, Greta Roberts will discuss results from a major 2012 study of analytics professionals that crosses industries, experience and skills. Practical insights shared include key best practices, trends and correlations that lend unexpected insight into building a strong and scalable analytic talent bench.
High-Performance Statistical ModelingThe explosive growth of data, coupled with the emergence of powerful distributed computing platforms, is driving the need for high-performance statistical modeling software. SAS has developed a series of procedures that perform statistical modeling and model selection by exploiting all of the cores available - whether in a single machine or in a distributed computing environment. This presentation describes the current capabilities of this software and offers guidance on how and when these high-performance procedures will provide benefits.
Medicare Age-InsPredictive modeling can be used to identify commercial members who are likely to leave as they age in to Medicare. The purpose of this study is to examine Excellus' commercial 64-year-old population in order to determine which factors play the greatest role in whether a member remains with us, and assist sales and marketing in targeting members that have a high risk of leaving in the hopes of intervening and retaining the member.
Using demographic, medical claims, pharmacy claims, product and member-specific data, we create a model to identify members who are at a higher risk of leaving Excellus as they age in to Medicare eligibility. The variable that proved to be most significant in predicting leavers was member persistency. Additionally, gender, MEDai score, physician costs, member cost share, pool area, and product class all were found to be statistically significant predictors. Based on the results of the study, our model correctly predicts roughly 59 percent of the leavers. The average annual premium for the 65-year-olds in the time period studied was approximately $4,500. Potential retained premium to the health plan from successfully intervening on even a small proportion of these "true positives" can be substantial.
This model will also be used to model expectations of a growing exchange/individual market.
High-Performance Analytics for Insurance and Risk ModelingThe insurance industry has the need to model the frequency and severity of adverse events each day. Accurate modeling of risks and the application of predictive methods ensure the liquidity and financial health of portfolios. Many times this modeling involves computationally intensive, large-scale simulation. SAS/ETS has specific procedures as part of its high-performance analytics suite to assist in this modeling. This talk will discuss the capabilities of HPCOUNTREG and HPSEVERITY, two SAS procedures which estimate count and loss distribution models in a massively parallel processing environment.
A New Face for Analytical ClientsSAS development teams are completing initial releases of the next generation of SAS analytical clients. Building on customers' favorite features with innovative enhancements, we offer a fresh interface for SAS Enterprise Miner and SAS Forecast Server. We will also be taking access to analytical procedures in SAS to the next level. Come see a preview of the latest build of these clients and a road map for future delivery of the analytical client user interfaces.
An Integrated Approach to Big Data Analytics from SAS and TeradataThe SAS and Teradata partnership was established in 2007 and since that time, numerous joint products have been released that capitalize on the strengths of each company. Those products leverage unique in-database and in-memory solutions and have been implemented in over 400 customer engagements. In addition to industry leading technologies, the partnership features a joint Center of Excellence and a joint support model to ensure effective use of those technologies.
This presentation will focus on the most recent joint product offerings and customer successes.
Advanced Illness Services: Enhancing Care at End of LifeHighmark's Advanced Illness Services/End-of-Life program, considered a best practice in the industry, helps members understand their medical conditions in order to make informed health care decisions through the support of uniquely qualified professionals who provide emotional support, facilitate decision making, prepare members for effective communication with their physicians, and arrange referrals to community resources. Using SAS Enterprise Miner to identify members who may qualify for this program, Highmark is helping their members achieve effective palliative care that provides the best quality of life in the face of advancing illness while respecting members' choices at the end of life.
SAS Recommendation EngineA recommender system generates meaningful recommendations for a collection of users for items or products that might interest them. Increased productivity and credibility and mutual benefit are common goals. Recommenders use neighbor methods, predictive models, heuristic search, data collection, user interaction and model maintenance. SAS is committed to delivering a robust in-memory recommendation system that supports ranking, similarity, prediction and classification applications. The SAS recommendation engine is being built on the SAS LASR Analytic Server to support large-scale item and user sets. The interactive IMSTAT is the SAS client tool that interfaces with LASR. The SAS recommender will also be fully integrated with SAS High-Performance Analytics.
Implementation of Predictive Model Factory for Analytics DeliveryThe Global Analytics department of Ford Motor Credit Company is responsible for developing and delivering predictive models for use in data-driven decision making. This presentation will focus on the effects of the model factory approach in leveraging economies of scale in data processing and shared learning, rapid response to business environment changes, adoption of new technologies and techniques in managing risk, and business globalization.
Use of Social Network Analysis to Analyze Physician Referral Patterns and Efficiency of CareAs the largest payer in the state of North Carolina, BCBSNC is looking at patterns of referrals across physicians to understand which relationships lead to higher-quality and more cost-effective care. Physicians often do not know when and how their patients access care through other providers in the health care system. Social network analysis (SNA) techniques are helpful at identifying how patients interact with the health care system, which providers share patients, and which formal and informal networks of providers manage patients effectively. We'll review a case study of how BCBSNC is using SAS tools to conduct SNA in order to identify opportunities to improve patient care.
Analytic Model Deployment: The Moment of TruthBuilding analytic models for the sake of identifying analytics-driven business insight is a worthwhile exercise, but the moment of truth happens when the resulting models are used to drive a change in customer behavior. To maximize the business value from analytics, organizations need to develop robust deployment practices that get models rapidly into the hands of staff interacting with customers in that moment of truth. Business processes supporting governance and transparency need to be developed so that the organization understands how model works, who built it and approved it, who in the field should use it, and which decisions it should inform. Organizations also need to manage their model portfolio as it ages so they are calibrated, enhanced and possibly replaced as their value decays.
This presentation will provide concrete steps for increasing the value from your analytics activities by actively operationalizing, cataloguing and tracking your portfolio of analytic models. You'll hear practical advice and lessons learned based on Way's experience with financial services companies that have successfully brought analytic models into the field by supporting their governance and monitoring their performance.
Customer Service Analytics in a Big Data EnvironmentThis presentation discusses the challenges T-Mobile faces to develop a big data analytics environment that integrates many disparate data sources into one cohesive whole. The presentation provides examples for successes and failure in building a big data platform as well as illustrations of the customer service insights made possible through big data analytics.
SAS High-Performance Analytics: Big Data Analytics for SAS AnalystsSAS High-Performance Analytics software exploits modern computer architectures by using a combination of parallel processing, in-memory analytics and distributed computation to enable you to mine your largest data and develop predictive models. This presentation describes new procedures and architecture options that support a wide spectrum of big data analytics. You will learn how this technology fits into your computing infrastructure and data centers, whether you use SAS data sets on a single server or you analyze massive data housed in enterprise data warehouses or Hadoop clusters.
A Framework for State Estimation Under Complex Influence Using SASIn the market, it is challenging to achieve accurate state estimations when business rules and environmental factors are changing. This is especially true when there is a high amount of dimensional data, limited sample size, and the change pattern cannot be captured easily via classical statistical methods. We propose a framework for solving this problem by using SAS to integrate a stochastic process module and a control module. A case study of this framework for a multiclass inventory state estimation problem has shown its effectiveness.
The Aesthetics of AnalyticsWe are encouraged to develop our understanding of the soft side of analytics - including the presentation skills required to present results and the people skills to understand the impact of change. As a step in developing those skills, it is helpful if we first appreciate for ourselves as analysts the aesthetic value in our work.
This presentation is inspired by The Architecture of Happiness, a book with the simple premise, "What is a beautiful building?" In the book we recognize that a large part of the aesthetic is the building's usefulness. With this inspiration we look for what is beautiful to us in analytics. Recognizing and relishing the aesthetic parts of our work is a foundation for sharing analytical inspirations with others. While we should not expect outsiders to appreciate this beauty as much as we do as practitioners, our recognition of the art of our work is a first step in gaining an understanding of our findings from others.