• Home

• Overview

  • About Analytics 2013
  • Pricing
  • Who should attend?
  • Why attend?
  • Join our interest list

• Pricing

• Speakers

  • Keynote speakers
  • Session speakers
  • Conference chairs

• Sessions

  • Agenda (.pdf)
  • Keynote abstracts
  • Session abstracts
  • Special events

• Training

• Sponsors

• Hotel & Travel

  • Hotel & Travel
  • About London

• Register Now

Sessions

Session Abstracts

Predictive Modelling   |   Text Analytics   |   Forecasting   |   SAS Presents   |   Analytics for Customer Insight
Analytics for Fraud   |   Analytics for Risk   |   Supply Chain Analytics   |   Introduction to Analytics

Track: Predictive Modelling

A CI solution with a data mining focus is helping Odense Municipality divide its citizens into different health groups to reduce and prevent long-term sickness Forecasting the caseloads and dynamics of universal credit is an analytical challenge that requires an integrated forecasting model for income-related benefits, housing benefits and tax credits. Fortunately, the Department for Work and Pensions already has the INFORM dynamic microsimulation model, developed by the Model Development Unit, which includes the income-related benefits.

Extending INFORM to include tax credits requires us to model a series of major tax credits policy changes. Full modelling of all variables required for direct adjustment for the policy changes within the forecasts was unfeasible in the timescales required for delivery and undesirable for model run times.

The solution: perform static microsimulation on the historical data to generate "counterfactual" entitlement under the successive policy changes. This gave us a historical series for each policy change, each with synthetic entitlement as if that policy had existed throughout the historical data. From these we could estimate regressions for our dynamic microsimulation using variables that we do model through the forecast period. The static microsimulation gives us more robust relationships between tax credits, additional benefit receipts and other characteristics after the policy changes than were possible from other simpler approaches. As a result, there were more robust forecasts of overlaps between benefits and tax credits, which are vital for modelling universal credit. The pharmaceutical industry is subject to various local pricing regulations for new drugs. Many interdependencies between launch markets mean the launch order can determine the price charged per dose. This pricing has an enormous effect on the total profitability of any new drug during the crucial patent-protected phase. With SAS/OR, pharmaceutical companies can juggle the various individual market regulations, pricing information and release schedules to simulate the optimal product launch sequence.

Track: Text Analytics

High-Performance Data Mining and Big Data Analytics is an upcoming book that describes the exponential growth of data and the latest trends in processing, analysing and using this expanding data volume. This presentation discusses the key concepts and factors to consider in storing, processing and operationalising your big data assets, both structured and unstructured. It also discusses the analytics life cycle, techniques for preparing and modeling big data, applications in big data analytics that you can apply in your company, and best practices that will help you get the most value out of your data. The book's authors share their extensive experience, as well as the experiences and success stories of leading companies in several industries. SAS Visual Analytics is a powerful tool for exploring big data to uncover patterns and opportunities hidden with your data. The challenge with big data is that the majority of it is unstructured data, in the form of customer feedback, survey responses, social media conversation, blogs and news articles. By integrating SAS Visual Analytics with SAS Text Analytics, customers are empowered to uncover patterns and relationships within both structured and unstructured data. This combination creates an enhanced view of your big data, enriching and visualising your data with customer sentiment and categorical flags, and uncovering root causes that primarily exist within unstructured data. This paper highlights a case study that provides greater insight into big data, demonstrates advanced business intelligence reporting, while enhancing time to value by using SAS Visual Analytics high-performance, in-memory technology along with the advanced capabilities of SAS Text Analytics. A big pharmaceutical company in Belgium applied business analytics across departments by integrating data from global sales operations, finance and supply chain. It initiated an innovative Web-listening project using text analytics to analyse patient comments available in social media. Doing so helps it reach its aspiration of being a patient-centric company that offers innovative new medicines and enables cutting-edge scientific research that is determined by the patients' needs. This talk will look at the three key elements needed to derive value from analytics across all the internal and external data available to an organisation. We will investigate three key areas for performance:

• Platform - choosing the right technologies.
• People - what to look for in successful teams.
• Process - driving toward an agile approach to analytics.

The world is changing at an alarming pace. The adoption of inexpensive and "always connected" technologies to communicate and share data is almost universal. In an age of ever-increasing change and complexity, organisations will need to sweat their data holdings and convert them into outputs and visualisations that matter. The appropriate collection and identification of the "right" material can make a positive difference, but this is no trivial task. This paper describes the current and future challenges in the collection, assessment and analysis of data to create actionable insight. The paper also discusses how approaches to information and data have changed through time and discusses some of the characteristics for today's organisations as they evolve.

Track: Forecasting

Escapo is the main distributor of drugs in Belgium for the CM Group. It provides nearly 100 pharmacies with the necessary stock. The pharmacies and 21.000 products (£13 million in stock and 20.000 orders per day) were wasting about £200.000 and close to £400.000 in dead stock per year. The objective of this project was to reduce the waste while guaranteeing a desired service level at all times. A new customised inventory management tool was developed for the automated optimisation of inventory levels, which included Escapo's central stock and the surplus stock at pharmacies. The results exceeded expectation, reducing stock by 35 percent and lowering the number of logistic movements between warehouses and pharmacies. A representative of Escapo shares his views on the project and the partnership with 4C Consulting in a customer testimonial. (See 4cconsulting.com/cases.) This paper takes the audience through the journey RWE npower embarked on with its Phoenix programme, that addressed internal and external challenges faced in accurately forecasting UK energy demand. The programme led RWE npower to implement a new forecasting platform that could provide greater flexibility and be more responsive to changing market conditions. This paper outlines the approach taken, technical capabilities implemented and benefits gained with the new forecasting platform. SAS Forecast Server offers a new, innovative process for automatically combining forecasts. Forecast combination, also called ensemble forecasting, is the subject of many academic papers in statistical and forecasting journals; it is a known technique for improving forecast accuracy and reducing variability of the resulting forecasts. By integrating these methods into a single software system, SAS Forecast Server surpasses the functionality of any existing software system that incorporates this capability. This paper describes this new capability and includes examples that demonstrate the use and benefits of this new forecast combination process. With more data being gathered at ever shorter time intervals, forecasting thousands of time series at an ever-increasing frequency has become a reality for many businesses. As time series data sets converge in size to those regularly faced by data miners, can forecasters learn from best practices to use algorithms and tools under development in predictive analytics? This presentation demonstrates how to use time series clustering and classification for a variety of situations:

• To identify most-similar and outlier motifs in hourly call centre data and electricity load data.
• To forecast new products by automatically finding similar past product launches.
• To identify suitable forecasting hierarchies beyond sales or assortment structures.

With similar algorithms recently pioneered in SAS time series explorer and available in SAS Enterprise Miner, all business cases can be directly implemented.

Track: SAS Presents

Big data has become the hot topic for analytical practitioners today. You cannot pick up a publication or research report without reading about big data. Analytical teams are being asked to develop hundreds, if not thousands of statistical models that take advantage of all this data. Organizations that are able to 'bottle up' and operationalise their analytical solutions experience significant benefits in terms of return on investment (ROI), enhanced productivity, and increased accuracy. Learn how SAS Model Manager and the SAS Scoring Accelerator can be used to automate the management, publishing and scoring of models, with results that can be seamlessly integrated into production processes. Revenue management (RM) applies to a wide variety of industries. Starting in the airline industry, it quickly spread to other travel and transportation domains, ranging from cruise lines and hotels to ticket prices and rental equipment. In recent years, it also gained a foothold in manufacturing and distribution industries. SAS has launched a product that can be custom-configured to cutting-edge revenue management solutions supporting any industry, as well as individual company policies. This product, SAS Revenue Management and Pricing Optimization Analytics, consists of a coherent set of modules based on SAS products geared at the revenue management technologies. In this talk, with a focus on business framework development, we demonstrate the tremendous benefit that SAS Revenue Management and Price Optimization Analytics can bring to companies in various industries. Using real examples of the solution's implementation, this presentation highlights how different business practices for RM can be incorporated with a glass-box approach. This presentation picks up where two previous talks - "Consequences of Poor Data Quality on Model Accuracy" at Analytics 2012 in Cologne, and "Data Quality for Analytics and the Consequences if It Is Not as Good as You Thought" at Analytics 2012 in Las Vegas - left off. This new presentation gives a detailed view on missing values in analytics and discusses their background, as well as methods in SAS for detecting and treating them. Learn how to handle the missing values in data mining and time series analysis and see how SAS is perfectly suited to assist in this task. This talk explains missing values from a univariate point of view as well as analyses them in a multivariate way. It also explains how to detect a pattern of missing values in analytic data and how to handle missing value in time series data. The content of this presentation was previously presented at the SAS Club Austria 2012 (Austrian SAS users event) as well as at the Austrian Statistics Week 2012 at the Vienna University of Technology.

Read this blog for more information about this presentation.

Track: Analytics for Customer Insight

British Airways (BA) needed a better understanding of the effectiveness of its marketing to know which types of campaigns are most effective and whether it would be beneficial to change the media mix. This presentation outlines how the OR department at BA developed an econometrics model in SAS to inform this strategy. The SAS France Academic Department reflected on how it could improve business analytics projects, particularly the ones using big data. These improvements were motivated by some misunderstandings around analytics factories, specifically how to industrialise the analytical process when there are different expectations for the outcome. What is the purpose of an analytics factory: developing IT roles, developing functional/business roles such as risk or customer intelligence, or developing analytics roles? The SAS Spring Campus is an innovative training program for master's degree students based on the three axes listed above. This presentation explains the best practices for developing the culture of an analytics factory, and examines the first session of SAS Spring Campus 2013 and its 20 master's degree students. A recommender system generates meaningful recommendations for a collection of users for items or products that might interest them. Increased productivity, credibility and mutual beneficial proposition are common goals. Recommenders leverage 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 SAS LASR Analytic Server. SAS Recommender will also be fully integrated with SAS High-Performance Analytics.

Track: Analytics for Fraud

Analysing tax returns, verifying import and export customs declarations, consolidating and checking the data from dozens of local tax collector's offices – these are just a few of the examples of data mining at the Belgian Federal Public Service (FPS). But apart from handling fraud detection and tax collection, those analyses lend themselves perfectly for improving FPS' customer relationship management with its citizens and for creating statistics, forecasts and simulations regarding all fiscal earnings. Specific analyses will enable us to recognize our stakeholders and enhance our service to citizens and companies. Our strategy is evolving toward a more comprehensive and shared approach that reflects the various risks; it is primarily aimed at preventing abuses. After presenting the general vision, the business case "VAT carousels" is explained. From an academic perspective, credit card fraud detection is a straightforward data mining problem, but in practice, it is a very complex problem. This presentation examines a few of the difficulties:

• Fraud detection must be cost-sensitive.
• Very, very few of the enormous number of transactions are actually fraudulent.
• Models must be interpretable, meaning black-box models are out of the question.
• Responses are needed in milliseconds so the implementation has to be analysed from the beginning.

Revenue, the Irish Tax and Customs Authority, has been using the power of data mining techniques that put 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 that scores case risk in 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, which is consolidating and improving its use of analytics. This paper describes Revenue's experiences using analytics, highlighting successes, lessons learned and possible future developments. The game has changed since the onset of the financial crisis. Governments aiming to reduce budget deficits can only deliver so much through spending cuts. It is now even more vital that tax agencies ensure individuals and businesses pay the tax they owe, and that welfare fraud and error are minimised. Pretty will explain how he helps tax and welfare agencies tackle noncompliance, evasion and error. He will share client stories where billions of euros were saved, generating a return of at least 25 times the original investment.

Track: Analytics for Risk

Many banks have identified balance sheet optimisation as a key strategic focus area. Efficiently managing a bank's balance sheet while maximising returns – and at the same time taking into account conflicting goals such as minimising risk within regulatory and managerial constraints – is a complex task. Using a trial-and-error approach can only deliver suboptimal solutions. Given the cost of capital, the total capital available and bank management's risk appetite, managers need to determine whether there exists an "optimal" balance sheet composition of assets and liabilities that will enable their organisations to achieve their strategic goals. Taking the current balance sheet as a starting point, this paper proposes a multiperiod, multiobjective approach to move from the current balance sheet to the optimal balance sheet, while taking regulatory capital limits into account. Basel III has introduced new balance sheet ratios and constraints. This paper reviews some of the methodologies and shows the implementation details for an optimal balance sheet model, which includes regulatory constraints on ratios such as capital ratio, leverage ratio, liquidity coverage ratio (LCR) and net stable funding ratio (NSFR) using PROC OPTMODEL. Based in Vienna, Raiffeisen Bank International has subsidiaries across eastern Europe. In 2011, the bank began using SAS Model Manager to create an inventory of different models across its subsidiaries and to monitor their quality. In a first phase of the project, four countries were selected as pilots and included into the model manager. The second phase added the larger countries, which had more than 20 models per country. This presentation outlines some of the bank's challenges and experiences in collecting model information and scoring data from different countries. Managing earnings volatility is an important component of firmwide risk management and is a regulatory requirement for banks. However, calculating earnings volatility and determining its allocation to business lines raise complex modelling and data challenges for large portfolios. SAS software provides an appropriate solution to these challenges. An implementation for a large banking portfolio is presented. Economic capital is calculated based on risk data for a portfolio of 1 million corporate credit facilities, through Monte Carlo simulation across several million scenarios in a grid computing environment. Risk measures are coherently attributed to business lines and integrated with financial reporting data, enabling economic capital-derived volatility metrics to be used to determine portfolio performance. This presentation covers:

• Integrated operations – from devices to more math and statistics.
• Risk-based prioritising and change of priorities over time.
• Determination of work according to priorities and constraints.
• Mathematics.
• Business-related problems explaining the model and results.
• Earnings, specifically improvements and positive results.
• Improvements and the reasons they are not implemented.

Track: Supply Chain Analytics

District nurses play a vital role in providing care in the community; however, their numbers are declining. The provision of nursing in the community is coming under ever-increasing demand due to the government's drive for a shift in care practices at a time where more people live at home with complex, long-term conditions. Working closely with a district nursing service in Wales, a study has been undertaken to determine the optimum skills mix of district nursing teams, so that nursing skills can be better matched to patients' needs. Statistical analyses of large healthcare data sets from the PARIS database provided insights into the district nursing service, allowing operational research techniques to be used to investigate a restructuring of the service. This paper documents how to improve the order distribution for a logistics service provider that accepts orders for fast-moving consumer goods from retailers. Due to daily fluctuations in orders, the logistics provider must have the maximum number of trucks needed on the days with the maximum orders, but this results in idle trucks on other days. By performing data analysis of the orders from the retailers, the inventory ordering policy of these retailers can be inferred and new order intervals can be proposed to smooth out the number of orders to reduce the total number of trucks needed. The result: the total number of trips made was reduced by an average of 20 percent. Complementing the proposed order intervals, the corresponding new proposed order size is computed using a moving average taken from historical order sizes, which satisfied the retailers' capacity constraints within reasonable limits. These insights were obtained and new solutions were proposed by integrating data analytics with decision analytics, thus reducing distribution costs for the logistics company. The study used Base SAS and Google Earth. SAS Retail Forecasting is a sophisticated solution that predicts consumer response to price changes, promotional events and marketing activity to generate a demand forecast at the store/SKU level. By applying advanced analytics to determine the net effect of promotions and price changes on whole categories, SAS Retail Forecasting evaluates cross-effects between products, forecasts new product sales and accounts for lost sales to generate a demand forecast. As a result, retailers improve their in-stocks and reduce out-of-stocks. This SAS solution provides advanced predictive analytics to generate a demand forecast for replenishment that:

• Uses models developed specifically for retail products with variable sales patterns and considers how critical causal factors such as price and promotions affect demand.
• Provides retail-specific methodology that models the consumer decision process to address the impact of price and promotion along with marketing and operational activity.
• Models all types of retail products from grocery to fashion, short or long life cycle.
• Accounts for seasonal effects and stock-outs.
• Helps shape demand by using future price, promotion and marketing activity.
• Is high-performance-enabled using threaded-kernel (TK) grid technology, which is scalable to meet demands of customers with different amounts of data.

This hands-on demonstration also shows how the solution integrates with the SAS Revenue Optimization Suite. With more than 100,000 regular personnel and an annual manpower bill in excess of £4.8 billion, the Army faces considerable challenges in exploiting the available data to support strategic decision making. This presentation covers how the Army is implementing a SAS based solution to address these issues, as well as improving data quality relating to its military manpower.

Track: Introduction to Analytics

What is the true value of statistical analysis, and how is it making a difference in organisations today? This session will delve into the world of statistics and demystify the concepts and techniques that are adding true value to organisations across multiple sectors.

This session will discuss a powerful, organization wide framework for statistical analysis, including: extensive data manipulation and sampling capabilities to prepare for analytic and modelling work; exploration tools such as association analysis and segmentation for better understanding large populations of customers; and statistical modelling techniques for predicting outcomes and reporting capabilities. The discussion will include use cases that show how companies are harnessing the power of SAS statistical analysis to realise true value through competitive advantage, reduce costs and better understand their customers. Building upon the concepts discussed in the value of the Statistical Analysis session, this How-To Session will demonstrate the real-world applications of SAS statistical analysis through practical examples.

This session will step through the layout and purpose of SAS Enterprise Guide and SAS Enterprise Miner as point-and-click interfaces to SAS statistical analysis. Emphasis in this session will be on the statistical capabilities of the two products, in particular, performing exploratory statistical analysis on data, conducting association analysis, building segmentation models and examining graphical diagnostics for model assessment.

At the end of this session participants will have seen firsthand how statistical analysis can be intuitively conducted within SAS Enterprise Guide and SAS Enterprise Miner and how key insights can be quickly determined from the data. Predictive modelling helps organisations answer the elusive question of "what will happen next?" Prediction involves building models based upon past experience to better understand, for example, whether a customer will respond to a marketing campaign. Using a targeted approach will reduce the cost of running a marketing campaign in the future by only marketing to a subset of our customer base.

In this session we will show how through the application of a wide range of predictive modelling techniques (including decision trees, regression modelling and neural networks) organisations can achieve value through more informed decisions.

This session will also define the value that can be achieved through forecasting. We will cover the benefits of delivering forecasts that reflect the realities of the business and improving the ability to plan future events with confidence. This session follows on from the Value of Predictive Modelling session by demonstrating how the techniques discussed would be implemented in practice. We will use SAS Enterprise Miner to demonstrate how a full end-to-end model development phase can be built using an intuitive point-and-click interface. SAS Enterprise Miner provides a guided mechanism to enable data visualisation, manipulation, querying, reporting, predictive modelling and model assessment through the use of a SEMMA (Sample, Explore, Modify, Model and Assess) methodology. Using a wizard-based interface that gives access to a wide range of data visualisation and predictive modelling techniques, users are guided through tasks and the required data, with charts or reports being created in a greatly reduced time. Building upon this we will also demonstrate how SAS Rapid Predictive Modeler provides business users with easy-to-use capabilities that allow them to quickly generate predictive models.

Finally we will show how SAS Forecast Studio can be used to generate large quantities of forecasts quickly and automatically without the need for human intervention unless so desired. We will show how the software automatically chooses the most appropriate forecasting model, optimises the model parameters and produces the forecasts. What is optimisation, and how can it add value to my current analytical processes? In this session we will show through use cases how SAS optimisation techniques are applied to solve real-world problems and add extensive value to the organisations that implement them.

The pinnacle of the use of analytics is the use of optimisation analytics to deploy resources appropriately to achieve the greatest usage within defined business constraints. It enables business to answer questions such as "what is the best that can happen given a set of constraints?" or "what loan rate can I offer to maximise the number of loans sold?"

The concepts covered in this session include mathematical optimisation, linear programming, inter/mixed-integer programming and nonlinear programming. In this how-to session on optimisation the concepts covered in the Value of Optimisation session will be put into practice. Through demonstration on real-world problems we will show, for example, how simple linear programming problems can be solved using the OPTMODEL procedure.

Optimisation has often been described as prescriptive analytics and as the top level on the data and analytic continuum. Despite the perceived and demonstrated value that it can bring, it remains an underused analytic capability in the SAS Analytics arsenal. Part of the problem is because of the perceived and real complexity in operations research problem formulation and execution. However, this complexity can be masked to the business user by presenting optimisation algorithms via simplified, ready-for-business interfaces.

Attendees to this session will come away with key insights into how optimisation routines are constructed and applied to real business challenges.

• examples of business analytics
• the analytics process model
• predictive versus descriptive analytics
• analytics model requirements
• post processing

• types of data sources
• sampling
• missing values
• outlier detection and treatment
• categorization
• weights of evidence coding
• information value

• target definition
• regression
• logistic regression
• decision trees
• regression trees
• evaluating classification models
• ROC analysis
• lift curve
• regression diagnostics
• case study: churn prediction in a telco context

• association rules (support, confidence, a priori, interestingness, and so on)
• cross selling and market basket analysis
• recommender systems
• sequence analysis
• segmentation
• hierarchical versus non-hierarchical (for example, k-means) clustering

• social network applications
• social network metrics
• social network-based inference
• Markov property
• relational logistic regression

• analytics model requirements
• model interpretation
• monitoring analytical models
• backtesting
• benchmarking
• data quality
• corporate governance and management oversight