Abstracts

In response for the need for precision during public investment projects, the Italian Ministry of Economic Development designed an application that estimates the length of a project's five main implementation stages. While VISTO is built mainly for administrations involved in investment planning, it can be used in general as a transparency tool during the public investments implementation process and can provide assistance in the selection phase when new projects compete for funding. VISTO is a Web-based application available both as a SWF file and HTML page. In this talk, we will elaborate on some recent challenges that have emerged when applying data mining within the financial services industry for applications such as credit risk management, fraud detection, and anti-money laundering, We will first discuss issues related to data quality and master data management. Next, we discuss some important technical data mining challenges such as model interpretability, model compliance, and learning using networked data. This will be followed by an overview of model monitoring, model back testing, and model benchmarking. Finally, we also discuss how to incorporate macro-economic effects into data mining models by means of stress testing procedures. Throughout the talk, key recommendations and clear guidelines will be provided with respect to the challenges mentioned. Ideas presented in the talk can also have relevance in other data mining application fields, e.g. marketing, medicine, pharmaceutical, and governmental. Predictive risk management is a continuously updated process in which predictive analytics are used to make proactive decisions and mitigate loss. To create and manage a successful predictive risk management solution, an organization needs a clear vision and sophisticated predictive capabilities and tools.

Telecom operators face two key challenges. As a result of aggressive competition in the telecom industry, acquiring new subscribers is becoming more difficult. Also, the amount of bad debt is increasing as a result of economical fluctuation and ambiguity. Balancing these two issues is one of the most important credit risk problems in mobile telecom operators.

Formerly, organizations attempted to minimize bad debt loss and predict the risk of their subscribers by looking at behavioral credit scoring and application scoring based on data mining methods and infrastructure. That information was then analyzed and modeled using data mining techniques.

This presentation will describe the new business model of building a predictive risk management department and infrastructure, including an end-to-end daily behavioral scoring process, designing an effective GUI and updating the application scoring process. An aura of science fiction still surrounds artificial neural networks and methods of artificial intelligence, fuelled by movies such as Terminator and Star Trek since the early 1980s. The promise of intelligent, self-learning algorithms developed using the principles of information processing in the brain, capable of universal approximation and any kind of nonlinear prediction, sounds all too promising. However, neural networks are already being used today by industry leaders including NASA, Microsoft, Toyota, DARPA and (of course) SAS in mission-critical applications: from flying and landing jumbo jets, detecting unknown minerals in NASA's Mars Explorer and identifying fraudulent credit card transactions in real time, to anti-terrorism measures in facial, fingerprint and bomb detection at airports.

But how can the power of neural networks be used in something as "mundane" as time series forecasting? Can neural networks be enhanced to outperform established statistical forecasting methods? What applications lend themselves to neural networks, and where do they struggle against established statistical methods? In this session, we will show case studies of how to improve forecasting accuracy in retailing by including bank holidays, promotions and weather information, in banking for forecasting daily cash machine withdrawals, and how an international consumer goods manufacturer successfully automated forecasting parts of his assortment with neural networks.

Attendees will learn:

• The biological background of neural networks (and why they are only regression after all).
• How to apply neural networks to time series prediction.
• How to increase forecasting accuracy of neural networks by including additional variables.

Results of conventional outbound campaigns are diminishing and marketers are shifting their focus to inbound activities. The active customer is getting the starting point of promotional activities, and the new sales force has to follow. To compete, you have to control these activities, for which you need numerous sophisticated models. In this presentation, we will show practical aspects of modern customer intelligence with the emphasis on the enhancement of the data set. A significant part of supermarket sales is driven by promotions. Delhaize has extensively analysed the impact of these promotions before, during and after the promotion period on the sales and margin of the affected products, as well as the other products in the category (e.g.: cannibalisation effects). The aim of the analysis is to create a promotion bible for product categories, with tailor-made recommendations on promotional best practices. The ultimate aim of the promotion bible is to drive the same amount of sales with a lower promotional investment.

This presentation will focus on the technical and analytic challenges in creating the promotional bible, as well as insights into the internal acceptance and utilization of the results at Delhaize. This paper introduces a new approach to build a graphical model with categorical variables (a free model) in the frame of structural equation models with an application on real marketing data. Advantages and drawbacks of this technique are presented and comparisons are drawn with other modeling methods, like Bayesian networks. Some potential research tracks are then outlined. Business changes quickly in the telecommunications industry, and it is essential to keep predictions up-to-date. Retraining and development of new models requires a significant amount of resources, and existing models are vulnerable to changes in underlying databases. This presentation demonstrates how to automate the predictive modeling process. It permits a multitude of models to be built and refreshed within a short time. This presentation will provide an introduction to the European liberalised gas market, specifically focusing on the challenges this market structure gives to an asset-backed trading company like DONG Energy. We will briefly explain how the business issues can be transformed into a mathematical optimisation model and how we can use the outcome of the models for commercial use. We will also address various technical aspects, such as selection of optimisation procedure and IT setup for daily batch execution. Today, much of the knowledge doctors have about the treatment path of a patient is covered in medical reports, such as documents describing the diagnosis and procedures followed. Very few clinical information systems today can process this knowledge without manually added parameters. Additionally, in most hospitals, DRG coding is a process regarded as necessary only for reimbursement purposes. This is why most hospitals do coding only on the basis of doctors' exit reports (manual work). In this way, neither the hospitalization process nor the quality and risk analyses of the patient's treatment can really rely on the knowledge contained in the doctors' reports. Before operation systems can be enhanced with this kind of knowledge, they must first be developed to be able to handle this information.

The described project enables the Luzern Canton Hospital (Switzerland) to deal automatically with medical knowledge from the beginning of a patient's treatment by making the data available in a structured way so that ICT systems can handle it as needed. With this view, the hospital has developed a solution together with SAS that "reads" doctors' reports and translates them into the necessary software codes for the system to operate as described above. This model has been enhanced for the entry report in order to generate a complete patient treatment picture describing and structuring the patient's path through the duration of hospitalization. Analytics are a key component of the SAS Business Analytics Framework. Since the beginning of SAS, a significant amount of effort has been invested in the development, enhancement and nurturing of analytical software products that help solve your business problems. Some of the key drivers for the development of new analytical methods and techniques have been the need for:

• More powerful algorithms to address scalability as well as performance
• More flexible models to address an ever increasing range of applications
• Better visualization techniques that are suited to the different analyses
• Easier deployment of complex analytical algorithms for wider use across the enterprise

This presentation will provide a perspective from R&D that describes the many ways in which we have responded to – and are continuing to respond to – the above needs by constant innovation in the depth and breadth of analytical tools and techniques. The true effectiveness of a marketing campaign isn't response rate! It's the "incremental" impact – that is, additional revenue directly attributable to the campaign that would not otherwise have been generated. Yet traditional targeting criteria are often designed to find clients that are interested in the product but would have bought it whether or not they received a promotion. In such cases, the incremental impact is insignificant and the marketing dollars could have been spent elsewhere.

Net Lift Models are designed to maximize incremental impact by targeting the undecided clients who can be motivated by marketing. These "swing customers" are akin to the swing states of a presidential election; data miners could learn a lot from presidential campaigns.

Beyond targeted marketing, Net Lift methodology delivers tremendous performance improvements for deployed churn models – retaining "savables" while avoiding the adverse "reverse" affects that retention outreach triggers for some customers – as well as other creative business applications.

This keynote will demonstrate how to build Net Lift Models (also referred to as Uplift or Incremental Lift) that optimize the incremental impact of marketing campaigns, discussing the pros and cons of multiple core analytical approaches. Organisations today are increasing their use of analytics to predict business outcomes more effectively, improve business processes and turnaround time, and ultimately increase profit. Marketing, credit scoring, risk, churn and attrition models, among others, play an important role in any organisation today, but developing and managing the increasing number of models in production and development environments can be quite a challenge. Financial industries today also need to show capabilities of model validation and governance as outlined by regulatory acts like the BASEL II accord and other governing bodies.

To succeed in this, analytical models need to be managed properly from the scoping stage to the production environment. Therefore, model lifecycle management is a key process that ensures an efficient and iterative process that can be easily monitored and validated at any stage.

This presentation will cover a few integral parts for the effective development and management of models including the software, infrastructure and lifecycle management as defined by the Nedbank BIS DSS team. The challenge of a telecom operator, besides recruiting new customers, is to reduce the attrition rate and at the same time maintain or increase the average revenue per user, all under the greater mission of running a profitable business.

Highly individualized communication reaches its targets with high precision via a wide variety of channels, e.g., Invoice, SMS, MMS, DM, TM and self-service. Besides increased lift values and higher yield on the campaigns, everything is now launched with less effort and more efficiently in an automated process.

This presentation will give deeper insight into how the Swedish branch of the global Telco 3 works with strategic customer lifecycle management (CLM) and the critical contributions from SAS Analytics and SAS Enterprise Miner in this process. Each day, nearly 30,000 aircraft take to the skies of Europe, moving 2 million people and tonnes of high-value goods. Achieving a safe, orderly and efficient flow of traffic, whilst minimizing the impact on the environment, requires a complex layered approach to management and planning on time scales from seconds ahead to tens of years. The job of EUROCONTROL is to coordinate air traffic management across the skies of its 38 member states. One of the ways we do that is to provide forecasts of future traffic.

For more than 30 years we have been providing annual flight forecasts, but in recent years the demands have grown for forecasting more frequently and at ever-finer resolution: by month, by week, by state, by flow, by airport, by aircraft type or by market segment.

We face challenges which many forecasters will recognize: how to manage the flows of data and the preparation of forecasts; how to monitor and maintain quality; how to benchmark and improve; and how to stay within what can be done and yet keep your customers, who want more, on board. This paper will describe how we are responding to these challenges with the help of SAS. Statistical knowledge and IT technology is crucial to succeed. Statistics Norway contributes to sustainable development of statistics in several countries. The company is helping to build statistical capacity through long-term cooperation with sister organizations. With the main goal of fighting poverty, SAS and Statistics Norway are working together to help developing countries access powerful tools to provide statistics for evidence-based policy making.

The presentation will give a brief orientation of the countries with which Statistics Norway is co-operating, the status of its work in these countries and how SAS is used in the process. Selecting customers for sales activities based on analytics is one of the most popular applications in analytical customer relationship management (CRM). However, customers are not solely waiting for their bank to contact them. First and foremost, they use ATMs, internet and telephone to handle their banking transactions, to gather product information or to get in touch with their advisor.

How could Erste Bank take advantage of its customers' activities and at the same time offer the right products and solutions to the right customers? Why don't we offer products our customers really need via those points of contact they actually use?

This presentation shows how accurate, individualized product offers for customers are determined and how the paradigm shift from a product-centered to a customer-centered view was implemented at Erste Bank. With the ever-increasing number of analytical models required to make fact-based decisions, as well as increasing audit compliance regulations, it is more important than ever that these models can be created, monitored, retuned and deployed as quickly and automatically as possible. This paper, using a case study from a major financial organisation, will show how organisations can build a model factory efficiently using the latest SAS technology that utilizes the power of in-database processing. This presentation provides an overview of the current state of technology in the magazine industry and offers objectives to develop and implement a new forecasting platform to help simplify workflow, guarantee continuity and reliability, and quickly adapt to changes in the market. Risk and critical success factors of the project will be discussed. A question-and-answer session will follow. Kraft promotes statistical thinking and best practices globally with the development of software tools that execute complex data handling and calculation. The choice of software platform is based on both software capability and customer requirements so as to determine the optimal cost/quality solution.

JMP software was selected as the platform to provide a consumer test evaluation tool for the global sensory/consumer function in Kraft. It replaces a set of Kraft-developed data analysis macros written in various other software packages.

JMP scripts combine data handling and calculation requirements with novel high-quality graphical output tailored for import into PowerPoint presentations. Products are assigned names, labels and multiple colour profiles that are automatically used in all graphical routines.

The consumer test evaluation tool has halved analysis time, significantly reduced training time, minimised calculation error, ensured/increased use of Kraft best practices and reduced software costs. Its success results from a close collaboration between Kraft UK Banbury and SAS Professional Services in Marlow.

The application is now deployed globally under international and US Master License Agreements, which enables flexibility of deployment to an evolving customer base. Data preparation for analytics is an essential task for successful analysis, data mining and predictions. The quality of the data mart and its preparation is one of the most important influence factors for model quality and analysis result.

Data preparation itself must not be considered a solely technical or coding task. Analytic data preparation also includes the business point of view of the underlying business question, as well as analytical and data modeling considerations. The business question, for example, may impose restrictions on which periods may be used as input data. Different analytical approaches require different analytic data structures.

Based on his book Data Preparation for Analytics, authored by Gerhard Svolba, this presentation shows how data should be prepared in order to meet business, analytical and technical requirements. For the case of a one-row-per-subject data mart, it will show which aggregations will most effectively transform the customers' behavior into derived variables. Business analysts often need to develop campaigns based on predictive models to select which customers to target. The models often need to be developed in a short time frame without relying on a statistician. For these cases, SAS has developed the Rapid Predictive Model (RPM) extension for use by customers of SAS Enterprise Miner. RPM runs as a customized task in either SAS Enterprise Guide or using the SAS Add-In for Microsoft Office. It will automatically treat the data to handle outliers, missing values, rare target events, skewed data, collinearity, variable selection, and model selection. The results are presented in business terms as a simple-to-understand scorecard. The final model can be registered to SAS Model Manager and deployed to either a SAS server or through the SAS Scoring Accelerator to a relational data base. The presentation will show both the implementation and results from real-world trials. In regard to data analysis, the advantages of graphic display have been highlighted by several authors. Currently, large-size data sets can pose a challenge, as they can include a high number of lines or columns with complex categorical identifiers; relationships between variables can be subtle, in the spatial or temporal dimensions.

The paper highlights real-world examples from the oil industry of the advantages and performances of interactive solutions designed by SAS for exploratory data analysis: SAS/INSIGHT, SAS Stat Studio, and SAS Visual Data Discovery. Modern marketing concepts live by the quality of good scores, which precisely reflect the needs of customers. Vodafone Germany's score models are created automatically with batch data mining. The results are high-quality forecasts for campaign planning and control. In addition, optimised marketing budgets in the multichannel mix are used to automatically respond to market changes.