I’m looking forward to seeing the results. Feel free to comment this post with some remarks if needed.

• Flex and SAS – A Compelling Combination (Guillaume Curat, SAS Switzerland)
• Fraud Preventation at Viseca Card Services (Marcel Bieler, Aduno Group)
• Grid Computing with SAS (Guillaume Curat, SAS Switzerland)
• State of the Art Analytics (Udo Sglavo, SAS Global Practice)

One good aspect of this event was the possibility to meet a lot of people from SAS that we usually only know by their names. For example, there were people from SAS Education and SAS Support. It was very interesting to meet SAS team since I interact with them in my everyday work. The SAS Switzwerland team was really nice. In my point of view, the only drawback was the high number of presentations. Indeed, for people like us, with a stand, it was quite difficult to meet people that were not working for SAS since most of them were attending the talks. If you attended this event, feel free to comment this post.

You can access the pdf of the presentations here (thanks to @cu8sfan for the link).

First, I briefly define an interest group (IG) as an ensemble of visitors with the same interests. Each URL of a given website can be mapped into an interest. Examples of interests are Auto, Lifestyle, Entertainment, Sport, etc. A visitor can belong to zero, one or several interest groups at the same time.

Here is how the process works. First, a set of identified visitors (1) are tracked on the website. Since the pages they visit are recorded, one can deduce their IG. We build a model using as input the CRM data of these visitors and as output their IG (binary variables for each IG). In our case, this model is trained using a decision tree algorithm. All clients of the company, in fact the ones that have never been on the website, are scored using the obtained model. It is thus possible to infer their IG before they visit the website.

One possible use of such score is to decide which ad, content or product to show to a new visitor (that is already a client of the company). It is also possible to use this IG information with another channel such as mail, e-mail or phone in a 1-to-1 marketing context. For example, clients having a high score for the sport IG may be contacted for a sport product, etc.

So, what do you think of this approach? Have you tried similar approaches or completely different ones? I’m looking forward to reading your comments.

(1) “Identified” means that these visitors are also client of the company. We can thus merge CRM data with web behaviors, for example in a “log in” area.

If you’re coming to Baden, please come to visit FinScore (stand #2), I would be very happy to meet you there.

Link to the SAS Forum Switzerland (only in German)

To break the suspense, the book is of excellent quality. It is well written and the concepts are clearly presented and structured in the book.

Here are the chapters of the book:

• Targeting Approaches: A Unique Element in Online Advertising
• Online Advertising Reach and Frequency Concepts
• Winning Strategies in Online Advertising
• Display Advertising Online: What to Choose?
• The Connection of Online Search and Advertising
• Online Shopping and Buying
• Advertising Personally: E-mail and Word of Mouth
• Futures

The first and final chapters are particularly interesting. In the first chapter, you will learn anything you need to know about targeted online advertising. The different targeting approaches are explained: demographic, contextual, behavioral, geographic, daypart, etc. In the final chapter, experts of online advertising give their mind about the future of this field.

The book is closed by a comprehensive glossary about online advertising. It needs to be noted since it is more than 30 pages long. As a conclusion, I would recommend this book to anybody working in online advertising or behavioral targeting. It is a must have in your library if you’re working in this field. Congratulations to the ARF team for this excellent book!

System identification involves identification of a behavioral model that best explains the measured behavior of a structure. This research uses a strategy of generation and iterative filtering of multiple candidate models for system identification. The task of model filtering is supported by measurement-interpretation cycles. During each cycle, the location for subsequent measurement is chosen using the predictions of current candidate models. In this paper, data mining techniques are proposed to support such measurement-interpretation cycles. Candidate models, representing possible states of a structure, are clustered using a technique that combines principal component analysis and K-means clustering. Representative models of each cluster are used to place sensors for subsequent measurement on the basis of the entropy of their predictions. Results show that clustering is necessary to identify the different groups of candidate models. The entropy of predictions is found to be a valid stopping criterion for iterative sensor addition. Clustering helps classify models and, thus, it provides useful support to engineers for further decision making.

Update: Due to the number of requests for the paper, I have put it on my website. You can now directly download it: Improving System Identification using Clustering (pdf)

• Data into results (Sébastien Derivaux): This new blog, written by a consutlant in data mining, has interesting content about data mining applications. Also some book reviews.
• AI, Data mining, Machine learning and other things (Markus Breitenbach): Markus writes about machine learning with a focus on statistics, security and AI.

Enjoy reading these blogs and feel free to share your own blog discorey by commting this post.

Nowadays, customer data (CRM) and web behaviors are not employed to enrich each other. This is far from optimal because, huge leverage effects can be derived from each of these data domains. Currently, CRM or simply customer data which is available in most large companies is not used to enrich and enhance web user interactions and vice-versa web behavioral user data is not used to enrich customer profiles. As a result, one end up with a limited customer profile, which enables only limited personalization of customer interactions.

There are two types of web site visitors: identified and anonymous. For identified visitors, both CRM and web data exist. For anonymous visitors, only web data are available. Usually, the majority of website visitors are anonymous. The principle behind the developed solution is to build an extended customer profile by merging CRM data about the customer and his web behavior. A subset of these visitors is used to train the model and, using this model, a score is computed for any other identified visitors.

With this extended customer profile (web behavior + CRM data), a series of recommendations as well as web interest affinities (e.g. sports, technology, product type A, product type B) can be generated for customers with repeated web site visits. I have used data mining techniques such as decision trees to infer recommendations. I have been able to obtain up to 390% increase in identified visitors click-through rate.

Up to now, the solution provides a unique score for each visitor. The next step would be to have N scores for the N different products. It would then be possible to have a prioritization among the products (the higher the score, the higher the priority).

Feel free to give your opinion about this project. Also, if you’re working on a similar solution, don’t hesitate to comment this post and share your ideas.

• From SAS to Java: In the first look it seems difficult to compare SAS with Java code, since they are very different (Java is web oriented). However, it seems that DullesOpen has a tool, named Carolina, to convert SAS into Java code. You can test their tool online and I have contacted them to get a trial version. I will let you know when I have more information about this tool.
• From SAS to R: SAS system and R language (the free version of S programming language) are quite close since they are both statistics oriented. Of course, the syntax is completely different. Up to know, I didn’t found any tool for automatic SAS to R code conversion. I will comment this post as soon as I will find some interesting tool.

Ajay from DecisionStats told me about MineQuest. It seems that they have a tool to convert SAS code to WPS (World Programming System), a SAS Base clone. If you have some suggestions or way of converting SAS to any other language, you’re welcome to comment this post.

• The Online Advertising Playbook: this is a must have book if you work in the online advertising domain. It clearly explains the different kinds of online advertising (demographic, geographic, behavioral, etc.). Many details about successful online advertising strategies are given such as impact and frequencies of ad display. I’m still reading the book and will certainly write a review when finished.
• Enterprise Knowledge Management: The Data Quality Approach: data quality is a key step to most important data mining projects. In his book, David Loshin explains all the details of data quality with clear explanations about the methodology to adopt.
• SAS Certification Preparation Guide: Base Programming for SAS 9: since I wanted to learn a bit more about SAS, I ordered the basic and advanced official books for the SAS exam. These are two really big books (with a lot of space to comment the text) in which every details of SAS are explained in a very didactic way.

Feel free to comment if you have remarks on these books or if you’re reading other interesting books.