San Francisco • Chicago • New York City
Berlin • Washington DC • London
Delivering on the promise of data science
Click here for upcoming PAW events

Agenda – London, UK – October 29-30, 2014

DAY 1: Wednesday 29 October 2014

9.30am-10.30am - Conference Room 3

Opening Session:
Le Mariage Parfait - Combining Logit and Ensemble Modeling for Increased Customer Churn Detection

In our fascinating domain of Predictive Analytics, logistic regression has been the workhorse for decades - both among practitioners and researchers. Analysts have praised logit models for their predictive performance and high degree of interpretability. More recently, we have seen a growing enthusiasm for ensemble-based techniques - supported by results indicating that performance gains result from combining the outcome of different classifiers. In this presentation, we explore new ways of improving the industry standard using ensemble-based logic - and we illustrate promising performance results on a test bed of customer churn data sets. Presentation based on a joint research study with Kristof Coussement (PhD), Professor of Marketing Analytics at IESEG School of Management, France

Speaker:
Dr. Geert Verstraeten, Managing Partner, Python Predictions

[ Top of this page ] [ Agenda overview ]


10.55am-11.40am - Conference Room 3

Case Study: Belgian Government
Gotch’all! Advanced Network Analysis for Detecting Groups of Fraud

The Belgian Social Security Institution is a federal agency that registers and monitors every active company in Belgium, and is responsible for the collection of employer and employee tax contributions. These contributions are collected at employer level, making this process highly sensitive to payment fraud. As traditional techniques fail to meet the complex requirements of fraud, advanced social network analysis (SNA) offers new insights in the propagation of fraud through a network. We observe that fraud is often not something an individual would commit by himself, but is organized by groups of people loosely connected to each other. In this presentation, we will discuss how network analysis can ameliorate detection models for (1) detecting suspicious individual behavior based on his/her relationships to others; and (2) uncovering the so-called webs of frauds, i.e. groups of people frequently associated with fraudulent activities. We apply and validate the performance of the techniques on a real-life data set provided by the Belgian Social Security Institution.

Speaker:
Veronique Van Vlasselaer, PhD researcher, KULeuven

[ Top of this page ] [ Agenda overview ]


11.45am-12.30pm - Conference Room 3

Internet of Things Meets Customer Intelligence

The Internet of Things (IoT), which excludes PCs, tablets and smartphones, will grow to 26 billion units installed in 2020 representing an almost 30-fold increase from 0.9 billion in 2009, according to Gartner. The Internet of Things is the network of physical objects that contain embedded technology to communicate and sense or interact with their internal states or the external environment. For this presentation, eight totally unrelated device data sources are pulled together to create new fact based insight about customers and their behavior. Many new techniques were tried to make this happen and those techniques will be summarized. The insights and learnings gained are fascinating. As with previous presentations of Phil Winters at PAW London, public data and readily available open source software are used, making the presentation not only practical but reusable.

Speaker:
Phil Winters, Senior Managing Partner, CIAgenda

[ Top of this page ] [ Agenda overview ]


1.40pm-2.40pm - Conference Room 3

Keynote:
The Improbability Principle: Why Coincidences, Miracles, and Rare Events Happen Every Day

The improbability principle says we should expect to see amazing coincidences and highly unlikely events occurring. The principle is not a contradiction, but is the consequence of five laws, all based on the solid mathematics of probability. The laws are explained and illustrated with examples ranging from lightning strikes, lottery wins, financial crashes, and extrasensory perception experiments.

Speaker:
Prof. Dr. David J. Hand, Professor of Mathematics, Imperial College, London

[ Top of this page ] [ Agenda overview ]


2.45pm-3.30pm - Conference Room 3

Case Study: Tata Sky

How Analytics Boosted Cross Sell Ratios for One of India's Largest Cable TV Companies

We built next product purchase algorithms that created a unique recommendation engine called Subscriber Preference Modeler (SPM) for Tata Sky, India’s largest cable company. These algorithms aim to cross sell bundles as well as a number of unique services such as gaming applications. We will describe how analytics & smart use of Big Data played a leading role in improving cross sell and reducing churn rates to industry best levels. We will articulate the techniques used on imperfect data and how they impacted business. Our session will describe how unique attributes of ethnicity and innovative data fusion techniques were used to build the SPM engine which has now become the principal cross sell engine for Tata Sky.

Speaker:
Ajay Kelkar, COO, Hansa Cequity

[ Top of this page ] [ Agenda overview ]


3.55pm-4.40pm - Conference Room 3

Case Study: Activision
Cheating Detection in Call of Duty

Call of Duty is the #1 video game on the planet... and some people cheat. When people cheat in the game (which is called 'boosting') it ruins the playing experience for other players. It also creates issues with our leaderboards, since people who are at the top of such boards might be boosters. We developed a detection system to catch these boosters that is similar to fraud detection systems used for credit cards and insurance. This session will cover how we internally sold the capability, how we built the system and the impact it has had on the organization.

Speaker:
Arthur Von Eschen, Sr. Director, Game Analytics, Activision

[ Top of this page ] [ Agenda overview ]


4.45pm-5.45pm - Conference Room 3

Predicting a Better World: Three Case Studies Using Data for Good

We all know (or think we do) that prediction can increase profits, help us to understand our customers, and drive businesses forwards. Can this work in the world of not-for-profits? The challenges are just as tough, the availability of good data and good analysts even tougher, but the opportunities to make the world a better place are huge.
In this session Duncan will look at three recent examples of how predictive analytics has helped change charities. These examples come from the work of DataKind UK, a charity that links analytics volunteers with organisations needing support.

Speaker:
Duncan Ross, Trustee, DataKind UK

[ Top of this page ] [ Agenda overview ]

DAY 2: Thursday 30 October 2014

9.30am-10.30am - Conference Room 3

Keynote:
The Peril of Vast Search (and How Target Shuffling Can Save Science)

It's always possible to get lucky (or unlucky). When you mine data and find something, is it real, or chance? The central question in statistics is "How likely could this result have occurred by chance?" Ancient geniuses devised formulas to answer this question for multiple special-case scenarios. Yet, their calculus only applies to quaint, handmade analyses, where only a few hypotheses are considered. However, modern predictive analytic algorithms are hypothesis-generating machines, capable of testing millions of "ideas". The best result stumbled upon in its vast search has a much greater chance of being spurious. Such overfit is particularly dangerous, as it leads one to rely on a model molded to the data noise as well as signal, which usually is worse on new data than no model at all. The problem is so widespread that it is the chief reason for a crisis in experimental science, where most journal results have been discovered to resist replication; that is, to be wrong!
The good news is an antidote exists! Dr. Elder will explain the simple breakthrough solution -- still rarely employed, though newly being re-discovered in leading fields. John will illustrate how to use the resampling method he calls "Target Shuffling" in multiple learning scenarios, from model fitting to data exploration, showing how it calibrates results so they are reliable - essentially providing an honest "placebo effect" against which to test a new treatment (finding).
Bottom line: Honest Data Science can save Experimental Science!

Speaker:
Dr. John Elder, Founder & Chair, Elder Research

[ Top of this page ] [ Agenda overview ]


10.55am-11.40am - Conference Room 3

The Power of Simulating Buying Flows

When making choices concerning the opening or relocation of their shops, retailers and project developers often trust their gut feeling. Although intuition and experience certainly serve as a valuable compass, wrong decisions are made too often. In this session, Dieter will present a better alternative: Simulating buying flows. This strong predictive model is a very advanced version of a gravity model commonly used and can be constructed to calculate the impact of opening, closing or remodelling shops. The outcome is not only the estimation of shop turnovers, but also the estimation of the future catchment area and the internal cannibalisation on other own shops. The concept has been tested successfully today for about 20 leading Belgian and international retailers (e.g. Media Markt, Decathlon, …) in almost any segment (telco, food, sports, DIY, Electro). The application possibilities go much further than one might think at first.

Speaker:
Dieter Debels, Founder, Geo Intelligence

[ Top of this page ] [ Agenda overview ]


11.45am-12.30pm - Conference Room 3

Case Study: Booking.com
User Engagement at Booking.com through Topic Modelling in Travel

Booking.com engages its users in different ways, like for example email campaigns or on-site recommendations, in which the user receives suggestions for the destination of their next trip. This engagement is data-driven, and its parameters emerge from the corresponding relevant past behavioural pattern of users in the form of collaborative filtering or other recommender algorithms. In this presentation we present a use case where a secondary database with meta-information about the recommended destination in the form of user endorsements was used to provide personalised recommendations. We model the endorsements using Latent Dirichlet allocation, a well-known principled probabilistic framework, and use the resulting latent space to optimise user engagement. We demonstrate measurable benefits in two distinct interactions with the user in the form of email marketing and menu-based website browsing.

Speaker:
Lukas Vermeer, Data Scientist, Booking.com

[ Top of this page ] [ Agenda overview ]


1.40pm-2.40pm - Conference Room 3

Keynote:
The Revolution in Retail Customer Intelligence

In this new era of Big Data, companies collect data in ever-increasing volume and variety. In the midst of Big Data, a revolution is taking place in digital customer intelligence. This session will describe the transition from reporting to data-driven decisions using predictive analytics now leading edge, but will soon become mainstream; success requires not only collecting data but also derived attributes. Case studies will illustrate shopping cart funnel management, shopping cart abandonment, marketing attribution, churn, and purchase propensity.

Speaker:
Dean Abbott, Co-Founder and Chief Data Scientist, SmarterHQ

[ Top of this page ] [ Agenda overview ]


2.45pm-3.30pm - Conference Room 3

Case Study: verivox.de
Use of Large Data Sets to Predict User Behavior and Optimize Marketing Spending

By processing the digital footprints left behind by users online it is possible to build statistical models which predict human behavior with surprising accuracy. With digital marketing budgets rising, the ability to deploy these resources efficiently and effectively is becoming ever more important. We will present a case study where a stochastic model has been developed to predict the visitor and transaction patterns on the website verivox.de with more than 90% accuracy. We will also show how this solution enabled Verivox to optimize its online and offline marketing budget allocation significantly. Verivox is Germany's leading consumer energy and telecoms price comparison website. In addition to emphasizing the business value of such a model we want to talk about insights that will help organizations to span a bridge between managers and analysts and thus help them to find a common language to describe and understand common strategic objectives.

Speaker:
Gergely Kalmár, Predictive Analytics Specialist, Webrepublic AG

[ Top of this page ] [ Agenda overview ]


3.55pm-4.40pm - Conference Room 3

Case Study: NSA
Analytics to Detect Malicious Use of Internet Anonymizers

Good Guys Can Finish First! Internet anonymizers are often used to enable: (1) online anonymity; (2) data exfiltration; (3) botnets; and (3) other malicious software, not to mention censorship resistance. This session will describe how anonymizers are used and which analytic techniques that are effective in characterizing their presence/use for malicious purposes. At the end of this session, users will have an increased sense of analytic empowerment (via specific analytic examples) when it comes to how to effectively characterize Internet Anonymity behavior. Specifically, how information is anonymously stored, shared, and used on the Internet.

Speaker:
Dr. Aaron Ferguson, Technical Director, Cyber and Information Analytics Office, National Security Agency (NSA)

[ Top of this page ] [ Agenda overview ]


4.45pm-5.30pm - Conference Room 3

Predicting the Future: The Data Mining Approach to Time-Series Analysis

Want to predict the future? It may be news to some that Predictive Analytics is not always about that! Yet sometimes it is: when we have a series of past events, and want to predict an event in the future, that is when we want to build predictive models of time-series data. Analysing time-series data is not new, but data mining has a special approach to this kind of analysis, based on the principles of the CRISP-DM methodology. Tom will explain how it’s done: how to go beyond trend analysis, from business understanding, through data preparation and modelling to putting the results of use in the business, and also explain why this approach can help you achieve your business objectives more directly than any other.

Speaker:
Tom Khabaza, Founding Chairman, Society of Data Miners

[ Top of this page ] [ Agenda overview ]

Share |

Bronze Sponsors 2014

Sponsor 2014

Media Partners 2014

​Blogpartners​

Check out the official PAW Blogpartners
© 2017 Predictive Analytics World | Privacy

Program by: Python Predictions
Python Logo

Produced by Prediction Impact, Inc. and Rising Media, Ltd.

Predictive Analytics Company           Predictive Analytics Event Producer