Randy Kerber
1422 Flora Avenue
San Jose, CA 95130
408-244-8147
408-621-3684 (cell)
randy@techstories.com
Experience
Medco Health Solutions, Inc.
November, 2004 to present
Technical Specialist. Responsible for application of semantic and artificial intelligence technologies to the configuration, management, and delivery of prescription drug benefit programs.
Consultant: Business Intelligence, Software Development, Knowledge-based Systems
February, 2001 to November, 2004
Data analysis and software development. Creation of a data mining tool for use by data mining practitioners, employing a data-flow model to integrate multiple data analysis components within a single Java-based visual programming environment. Implementation of an algorithm to discover relationships between survey responses in a survey database. Development of tools to extract corporate event information from a respository of text documents. Implementation of a web application for marketing professionals to create and distribute HTML-based electronic documents (newsletters, surveys, product info) and track responses in a prospect database. Use of metadata and ontologies for Business Activity Monitoring and data integration. Use of Protégé to create OWL ontology for product configuration.
MyRaptor.com
April, 2000 to November, 2000
Director, R & D. Responsible for development of a Bid Recommendation Engine, providing on-line customers with a strategy for bidding on airline tickets from travel auction websites. Bid recommendations were derived via a combination of encoded heuristics and patterns derived from collecting and analyzing airline fare and flight data.
Data Mining Consultant
September, 1999 to April, 2000
Data mining software development. Customization of data mining software and processes for an e-commerce company adding personalization and target-selling capabilities to their site management product.
NCR
Human Interface Technology Center, Atlanta, GA
June, 1995 to December, 1996 (contractor)
December, 1996 to December, 1998 (employee)
Data Mining Lab, San Diego, CA
December, 1998 to September, 1999 (employee)
Data mining process standard. Co-author of the CRISP-DM (CRoss Industry Standard Process for Data Mining) process model. The model was developed by a consortium of NCR, ISL (now SPSS), and Daimler-Benz, with European Union support, and disseminated via a series of European and American seminars for data mining professionals. For project information see http://www.crisp-dm.org. For the final report see http://www.crisp-dm.org/CRISPWP-0800.pdf .
Data analysis: retail promotions. Analysis of grocery store sales data to predict effect of various promotion types (e.g., coupon, in-store display, price reduction) and price points on sales volume.
Data analysis: telecommunications. Analysis of telephone company databases (for Telefonica de Argentina). Predict which customers are most likely to respond to add-on product offers (e.g., caller ID, call waiting, etc.). Developed customer-value model, to assess the return on investment of each individual customer. Developed software to detect which products correlate with the purchase of other products.
Data analysis: service contracts. Analyze database containing equipment service contract histories and the maintenance work orders to fulfill the contracts. Identify which service customers are most in danger of not renewing their contracts.
Active Templates. Development of Active Templates, a research methodology for performing and delivering data mining solutions. A completed template integrates, into a single data structure, the actions, results, and documentation produced during a data mining project.
Data mining tool evaluation. Performed a study of the data mining tool marketplace. Compiled a list of approximately 100 commercial and public domain tools, covering techniques such as decision trees, neural networks, rule induction, and genetic algorithms. The most promising tools were obtained and evaluated in greater detail. The objectives were to study the marketplace to assess whether NCR should develop tools, evaluate the appropriateness of data mining products for NCR's service customers, and identify promising companies for partnership and referral agreements.
Prototype development. Developed a prototype Case-Based Reasoning (CBR) system in Java. Though general purpose, the CBR tool was targeted for use in medical domains as part of the National Medical Practices Knowledge Bank project, an ATP project jointly funded by NIST, AT&T, NCR, Allegheny General Hospitals, and Carnegie Mellon University. The CBR system retrieved, from a database of patient records, cases that were similar to a physician's current patient case.
Statis, Inc.
Sebastopol, CA
December, 1998 to April, 1999 (contractor)
Data mining product prototype. Developed a software tool, in Java, to perform "Bootstrapping", a statistical analysis method for creating predictive models. This tool was the prototype of a phase I SBIR (Small Business Innovation Research) project.
Lockheed Missiles and Space Company
Artificial Intelligence Center, Palo Alto, California
August, 1986 to June, 1995
Data mining software research and development. Research Scientist. Designed and implemented a rule discovery algorithm that takes as input a database of instances whose outcome is known and automatically generates rules of the form "IF instance X satisfies criteria P THEN predict the outcome of X is C". These rules can then be used to predict the outcome of newly presented instances. This rule induction system was integrated with a database query tool to form the Recon™ data mining system, a commercial software product marketed to both government and private sector customers.
Data analysis. Used Recon and other data mining and statistical tools to analyze data for customers in a variety of industries: aerospace, stock market, sales/marketing, pharmaceutical, and financial.
Bayesian inference and discrimination. Used Bayesian (belief) network and Kalman filter technology to track and discriminate multiple vehicles moving across a region given intermittent and noisy sensor readings of the form "possible vehicle near coordinates (x,y) at time z".
Knowledge representation. Built ontologies (highly structured knowledge bases) using a complex knowledge representation and inferencing system called Loom, which supports the definition of classes, relations, and production rules in a predicate logic-like language.
Hughes Aircraft Company
Artificial Intelligence Center, Calabasas, California
May, 1985 to Sept, 1985
Research in the area of intelligent, cooperating, software agents.
University of Southern California / Information Sciences Institute
Marina del Rey, California
October, 1984 to April, 1985.
Implemented a transformational program that created synchronized parallel algorithms from a formal specification of desired input/output behavior and constraints.
Education
M.S. Computer Engineering B.S. Computer Science
University of Southern California, 1986 University of Minnesota, 1983
Research Publications
Kerber, R., Beck, H., Anand, T., Smart, B. (1998). Active Templates: Comprehensive Support for the Knowledge Discovery Process. In Fourth Intl. Conference on Knowledge Discovery and Data Mining.
John, G., Miller, P., Kerber, R. (1996). Stock selection using rule induction. In IEEE Expert: Intelligent Systems & their Applications, vol.11.5, Oct. 1996.
Simoudis, E., Livezey, B., Kerber, R. (1996). Integrating inductive and deductive reasoning for data mining. In Advances in Knowledge Discovery and Data Mining (book chapter).
Kerber, R., Livezey, B., Simoudis, E. (1995). A hybrid system for data mining. In Intelligent Hybrid Systems (book chapter).
Simoudis, E., Livezey, B., Kerber, R. (1995). Using Recon for Data Cleaning. In Proceedings of the First Intl. Conference on Knowledge Discovery and Data Mining.
Simoudis, E., Livezey, B., Kerber, R. (1994). Integrating inductive and deductive reasoning for database mining. In Proceedings of the AAAI-94 Workshop on Knowledge Discovery in Databases.
Kerber, R. (1992). ChiMerge: Discretization of numeric attributes. In Proceedings of the Tenth National Conference on Artificial Intelligence (AAAI-92).
Kerber, R. (1991). Learning classification rules from examples. In Proceedings of the AAAI-91 Workshop on Knowledge Discovery in Databases.
Barletta, R., Kerber, R. (1989). Improving explanation-based indexing with empirical learning. In Proceedings of the Sixth International Workshop on Machine Learning.
Kerber, R. (1988). Using a generalization hierarchy to learn from examples. In Proceedings of the Fifth International Conference on Machine Learning.
Patents
Tate, B.; Pricer, J.; Anand, T.; Kerber, R. SQL-based Analytic Algorithm for Association. U.S. Patent #6,611,829, awarded Aug 26, 2003.
Cunningham, S.; Kerber, R. Method and Apparatus for Optimizing Promotional Sale of Products Based Upon Historical Data. U.S. Patent #6,029,139, awarded Feb 22, 2000.
Simoudis, E.; Livezey, B., Kerber, R. Method for Generating Predictive Models in a Computer System. U.S. Patent #5,692,107, awarded Nov. 25, 1997.