San Francisco Bay Area ACM

Location: NASA Exploration Center, Moffett Field, CA

Date: TUESDAY November 10, 2009; 6:30 pm  Notice: date changed from Wednesday to Tuesday!

Cost: Free and open to all who wish to attend, but membership is only $20/year. Anyone may join our mailing list at no charge, and receive announcements of upcoming events.

Speaker: Khosrow Hassibi, KXEN

ABSTRACT

In last 20 years, the application of Predictive Modeling has gradually evolved and has become popular in many B2C companies. These applications range from offline targeted marketing to real-time credit card fraud detection. In these uses, an entity (typically a customer) is characterized by its static (demographic, psychographic, etc) and more importantly dynamic behavioral attributes usually derived from transactional data. These attributes focus on the individual customer itself. What if the customers interact? Customer interaction information could provide additional predictive power on customer’s future behavior. A typical example is Telco. Each Call Detail Record represents an interaction between two entities. There are implicit social networks buried in such data that if extracted could present another view of the customer behavior.  These social attributes can be used as predictor variables in a variety of predictive/descriptive models such as cross-sell, up-sell, churn, fraud, segmentation, etc. In addition to the use of this information in predictive modeling, such attributes can be used in understanding customer interactions and the communities in which they operate (application in viral marketing). The talk explains the challenges, the potential business value, and the process with preliminary results from a couple of case studies in Telco. By loosening the concept of interaction, such information can also be extracted and used from other data sources such as credit card data, buyer/seller data, payment data, etc with other business uses. The focus of the talk will be on real-world business applications and not on algorithms. The speaker will be interested in audience’s ideas and inputs on other potential uses.

BIOGRAPHY

Khosrow Hassibi is a Senior Technical Dierctor at KXEN

Khosrow Hassibi  is a Senior Technical Dierctor at KXEN, a data mining software vendor headquartered in San Francisco, CA. Dr Hassibi has a Ph.D. in EECS from Case Western Reserve University with special concentration in the areas of Intelligent Systems and Machine Learning. Dr Hassibi is also a graduate of UCLA Anderson Executive Program in Management. His expertise is based on seventeen years of design, development, consulting, and management in applying advanced data mining technologies to real-world applications in a variety of industries. He is known in the financial industry due to his contributions to payment card fraud detection technology at HNC Software where he was overseeing the design and development of real-time fraud detection analytics protecting hundreds of millions of payment cards worldwide. Dr Hassibi has worked with two of the early pioneers in the field of Neural Netwroks – Dr. Yoh-Han Pao (Founder of AI Ware) and Robert Hecht-Nielsen (Founder of HNC Software Inc now Fair Isaac Inc) – to develop solutions for hard real-world business applications. At KXEN, Dr Hassibi is responsible to support the sales team in all technical sales tasks such as customer technical assessment, proof of concepts, etc. He also provides input to product development for improvements on the products.

Prior to KXEN, he ran the R&D operations of Core Analytic Technology team at HNC Software Financial Solutions. The focus of his team was on development and improvement of real-time payment card fraud detection system called Falcon^TM.  The Falcon system is the de-facto payment card fraud detection standard and the market leader currently protecting 450 million active payment cards in 6 continents (80% market share).  During his tenure, he designed and developed a new generation of Falcon fraud detection models. This architecture improved potential fraud savings by an average of 30% translating in hundreds of millions of dollars of savings for credit card issuers. Falcon^TM is often mentioned as the most successful business application of neural networks technology. He has written a book chapter on the state of payment card fraud detection in the book titled “Business Applications of Neural Networks”.  Earlier at HNC, he was the principal investigator of one of the main DARPA/ONR research topics in Neural Networks – recognition of machine-printed cursive text. At the time in 1995 the final system he invented was chosen by US Government as the official OCR system for recognition of languages that use Arabic text. Earlier in his career at AI Ware, he was one of the two original developers of NNET^TM, one of the first multi-platform neural network modeling products.

20091110 – KXEN – Implicit Social Networks (presentation 1.6 MB PDF)