PLATFORMS: NASD Regulation, Inc., monitors stock market trades and quotes with its Advanced Detection System (ADS), which is comprised of KDD Explorer running on a Sequent NUMA-Q 2000 host and Sun SparcStation workstations.

BACKGROUND: The National Association of Securities Dealers, Inc., (NASD) is the largest securities industry self-regulatory organization in the United States. Through its subsidiaries, NASD Regulation, Inc., and the NASDAQ Stock Market, Inc., the NASD develops rules and regulations, conducts regulatory reviews of members' activities and designs, operates and regulates securities markets all for the benefit and protection of investors.

PROBLEM SOLVED: In order to maintain our leading position as a regulator of the securities industry for the benefit and protection of investors, NASD Regulation, Inc., needs the most effective automated capability to provide continued surveillance of NASDAQ and other stock markets. The market regulation department is responsible for monitoring of all activity to ensure compliance with market rules in order to provide a level playing field for all market participants and investors. An automated system is necessary to screen all trades and quotes according to approved, standardized criteria to identify potential violations for review and, if appropriate, regulatory action.

PRODUCT FUNCTIONALITY: The KDD Explorer toolset components provide a comprehensive environment for construction and operation of an integrated fraud detection and knowledge discovery system. Fraud detection capabilities are provided through rule and sequence-matching algorithms, which are used to detect instances in the database that match patterns that could indicate fraud. The toolset includes a user-interface written in Java that provides customized visualizations for depicting the relevant relationships in the data, as well as capabilities for multiple users to review and manage the alerts that are generated and the patterns that are used. Its knowledge discovery components consist of a decision tree module that extracts classification rules from the transaction data and an association rule module that extracts common co-occurrences of data.

STRENGTHS: KDD Explorer provides high performance, customizable data mining and fraud detection capabilities on large databases. It has a variety of flexible control structures. The algorithms are highly scalable and can be run in parallel to take advantage of multiprocessor platforms such as Sequent's NUMA-Q 2000. The user interface is developed in Java and runs on multiple platforms. Patterns are easily understood through expression in natural language.

WEAKNESSES: This KDD Explorer release requires extensive technical capabilities to be customized into an effective end-user application. It does not yet have data preparation capabilities allowing for sampling, selection, transformation or other operations on a data set. An integrated application requires extensive data preparation to define and implement appropriate attributes and representations.

SELECTION CRITERIA: The toolset was developed in conjunction with ADS for NASD regulation. Key requirements included combined discovery and detection capabilities, high performance on large databases, domain-specific visualizations and multiple strategies and representations for detection and discovery.

DELIVERABLES: NASD Regulation analysts use KDD Explorer for surveillance of the NASDAQ stock market for potential violative activity. Experienced analysts review these potential violations, or breaks, for appropriate regulatory action. ADS uncovers potential violations that would otherwise remain hidden in the data and allows analysts to focus their efforts on the data most likely to represent a real violation.

VENDOR SUPPORT: SRA and Sequent have both provided outstanding support to ensure that NASD Regulation's business needs are met effectively and rapidly.

DOCUMENTATION: The documentation is minimal at present. It consists only of developer's notes and slide.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access