Recent Applications Of

Recent Applications of KnowledgeBased System (KBS) Technologies

Tony Polito (E-Mail:), East Carolina University

Rik Berry (E-Mail: ), Morehead State University

Kevin Watson (E-Mail:), Marist College

Abstract

Despite earlier predictions for their demise, nowledge-based systems (KBSs) continued to mature throughout the 1990’s, finding application in a variety of business disciplines including finance, human resources, operatons and marketing. This paper identifies and discusses a number of recent specific applications of KBS technologies in industry.

Introduction

Only a few years ago, more than a few practitioners sounded the death knell for the field of knowledgebased systems, claiming it had not delivered on its earlier promises. Now, knowledgebased systems appear to be more generally recognized; ie, "entering the mainstream" of technology. (Newquist) This paper identifies some of the more recent specific applications of knowledgebased system technology that have garnered attention, and the functional areas they serve.

Human Resources

One of the areas with the greatest long-term potential usage of knowledgebased system is human resources, such as training and education. Neural net technology is founded in pattern recognition. Expect such technology to be used for diagnostic efforts; ie, to determine what learning is required, and to accept the role of a savvy and dynamic tutor, moving through quick feedback cycles, identifying areas for improvement and offering instruction in those areas. Raj Reddy, Dean of the School of Computer Science at Carnegie-Mellon University works in many areas, including "just-in-time" learning. He expects that in the future, due to training applications, every desk in the organization will require a PC (Anthes). Roger Schank at Northwestern University employs 150 scientists and graduate students at Northwestern University to offer and maintain a knowledge-based system for use in higher education. The "ASK" system facilitates such mass customization by allowing students to choose among predetermined lessons or releasing their own query on the system, thereby creating an electronic, "just-in-time" teacher (Ellis, Garfinkel). Smart One, a training software package that continually adjusts and customizes according to the skills of the user (Greenberg). Medical laboratories are also noted for using knowledgebased systems to train new hires and support onthejob problem solving (Roseman). Neural nets have recently found a controversial application in selection. The Chicago Police Department is using a knowledgebased system based on a neural net chip that identifies applicants with history indicating a risk for misconduct (Hoffman). The union has announced its intention to sue.

Finance

Some of the earliest applications of knowledgebased system technology was in the financial industries, which, with its highly intangible products, is especially well suited to its use. Here, application continues to grow. The Office of Technology Assessment has reported that use of knowledge-based systems can be employed to reduce money laundering by identifying suspicious patterns in money transfers; they seek to empower the Treasury Department to issue automatically approved subpoenas for bank records based on knowledgebased system identification (Anthes). GTE and others are using knowledgebased systems to identify abnormal calling patterns that may result from stolen cellular phones or codes (Adelson). American Express UK uses a knowledgebased system for application decisions. In addition to the incremental accuracy, the system offers the desirable characteristic of an objective explanation of rejection (Langley & Simon). Recently, MasterCard International announced that, in conjunction with scientists at Los Alamos, it has developed a neural net methodology for detecting credit card fraud (NYT). A number of other credit card issuers employ HNC's Falcon system to detect fraudulent patterns (Epper, 1994a). The Federal Home Loan Mortgage Association completes a credit check in a few seconds; the systems uses EDI to tap the requisite external databases that supply the information to the decision-making system. GE Capital use knowledgebased systems to assess loan risk as well (Marjanovic). Observers expect most home loans to be handled this way in the near future (Allen). Both Fidelity and LBS Capital Management are managing specific mutual funds using neural net technology (Star). Barron's recently advocates the evidence for, and the use of, knowledgebased systems for futures trading, arbitrage and portfolio construction (Laing). However, their exemplar neural net investment manager sent a false strong sell signal in mid-1995, anticipating a Dow of less than 4000 (Mandelman).

Production and Operations

Production and operations management is beginning to find general application for knowledgebased systems as well. TIPS, offered by i2 Technologies, is and acronym for Truly Integrated Planning System, used in manufacturing areas such as factory planning, advanced scheduling, distribution planning, and available-to-promise quantities. TIPS links to existing transactional databases for initial pattern recognition training (Vasilash). Singapore Airlines is implementing a knowledgebased system that manages the company's cockpit and cabin crew scheduling (Ying), a problem often handled through linear programming models in the past. The model operates in real-time, responding to sudden changes such as in-route illnesses. Westinghouse uses a knowledgebased system to predict production output quality, in this case being uranium pellets. The quality of pellets was married to manufacturing control settings, raw materials characteristics and other relevant variables on which a decision tree was based. Its usage led to increased throughput, higher output, reduced inventory and a $10 million increase in revenue (Langley & Simon). Enichem, an Italian chemical company, uses a knowledgebased system to support preventive maintenance scheduling. Measurement of electric motor vibration patterns provide the input for the maintenance decision made using a decision tree approach (Langley & Simon) A similar example can be found in the diagnosis and repair of helicopter blades (El Attaar and Hamery). Thomas and Kleiner also note its use in troubleshooting automobile repair. The Army uses a knowledgebased system for diagnosis for missile system maintenance (Keller & Knutilla). Pohang Iron and Stell in South Korea uses a knowledgebased scheduling system to receive raw materials into the appropriate stockpile and routing of materials from stockpiles into the crushing, blasting and blending shops (Langley & Simon). Medical laboratories are using knowledgebased systems for supplies purchasing as well (Roseman). HNC Software's Automated Real Estate System uses neural net modeling to appraise property values (Epper, 1994b). Kodak is using neural nets for general process and production control as well (Schwartz).

Marketing

The marketing and sales function is also finding application; first efforts focus mostly on efficiency gains. Recognition Systems, specializing in development of knowledgebased systems for marketing applications, offers software for lead generation that taps external demographic databases and internal customer records (Bird). A knowledgebased system called Territory Manager Workshop is being used to teach and test territorial management skills such as account evaluation and call route planning to pharmaceutical salesmen (Leaf). Barclays Bank PLC is using knowledgebased systems software to speed the processing of its credit card customer complaints (Iida). Examples of marketing-oriented systems that focus more on complex decision environments can also be found. Taco Bell now uses knowledgebased systems for site selection (Southerland). A knowledgebased system model for tourism management that controls such things as protecting the environment against objective of maximize growth has been implemented. (Curry). Marketing News' Howard Schlossberg advocates the use of neural nets towards "tougher problems" such as customer satisfaction and product positioning.

Product Content

Perhaps the most enormous potential for application is the incorporation of knowledgebased system technology directly into new products; here are some examples. GRC International uses a knowledgebased system to enhance their security system by allowing it to distinguish noise truly caused by intruders (Day). Merline Elettrodmestici, an Italian manufacturer is now incorporating a small knowledgebased system within its washing machine products that learn and accommodate the habits of the user. Interestingly, Otis uses a knowledgebased system to eliminate customer dissatisfaction with using their elevators; eg, estimation of how many people are waiting for elevators to optimize their paths (Chartrand, Corcoran). The concept of intelligent products, and products that contain intelligent agents, is limitless.

References

Adelson, A. 1993. Software sounds the alarm on cellular theft. New York Times. July 28: D5.

Allen, J. 1995. Mortgage machine OKs loans in minutes. Chicago Tribune. May 14: 16.

Anthes, G. 1995. Great expectations: Award-winning AI scientist Raj Reddy foresees a future of intelligent agents, voice input, and near-real-time data flows. Computerworld. 29(14): 89-91.

Anthes, G. 1995. AI system all washed up. Computerworld. 29(38): 12.

Bird, J. 1995. Logical guides to marketing. Management Today. 58-62.

Chartrand, S. 1993. Patents. New York Times. September 13: D2.

Corcoran, E. 1994. The rise and fall of a high-tech idea. Washington Post. July 8: A1.

Curry, B. 1994. Intelligent computer models for marketing decisions. Management Decision. 32(4): 3035.

Day, K. 1994. Security system can distinguish raccoons from robbers. Washington Post. May 24: D1.

El Attar, M. and Hamery X. 1994. Industrial expert system acquired by machine learning. Applied Artificial Intelligence. 8: 487-542.

Epper, K. 1994. HNC pushes neural network loan system. American Banker. October 13: 12.

Epper, K. 1994. First Chicago unit using credit card fraud detector. American Banker. April 22: 17.

Garfinkel, S. 1995. AI as training tool. Technology Review. August/September 1995. 98(6): 16-18.

Greenberg, I. 1995. Smart One tailors training to individuals. Infoworld. 17(4): 10.

Hoffman, T. 1995. Swamp thing. Computerworld. 29(32): 83-84.

Iida, J. 1991. Expert system to process complaints for Barclays. American Banker. February 8: 3.

Keller, B. and Knutilla, T. 1990. U. S. Arny builds an AI diagnostic expert system, by soliders for soliders. Industrial Engineering. 22(9): 38-41.

Laing, J. 1995. New brains. Barron's. February 27: 29.

Langley, P. and Simon, H. 1995. Applications of machine learning and rule induction. Communications of the ACM. 38(11): 54-64.

Leaf, H. 1995. The infallible salesman. Medical Marketing and Media. 30(8): 74.

Mandelman, A. Follow up: The computer is bearish. Barron's. August 7: 12.

Marjanovic, S. 1994. GE Capital Mortgage Corporation uses artificial intelligence system. American Banker. 159(204): 9.

Newquist, H. 1994. AI burnout. Computerworld. 28(16): 129-131.

New York Times. 1995. Mastercard sets fraud detection. May 10: D5.

Rooney, P. 1994. Smart agents serve users; neural network chip heralds artificial intelligence on desktop. PC Week. 11(24): 33-35.

Roseman, E. 1994. Using artificial intelligence in the laboratory. Medical Laboratory Observer. 26(5): 5759.

Schlossberg, H. 1994. Real hopes for research placed in artifical intelligence. Marketing News. 28(1): 8.

Schwartz, T. 1995. Applications on parade. Electronic Design. 43(16): 68.

Southerland, D. 1993. Cooking up software that tells where to put Taco Bells. Washington Post. July 27: D1.

Star, M. 1994. New AI contender looms; LBS goes aganist Fidelity in neural net mutual funds. Pensions & Investments. 22(21): 48.

Thomas, V. and Kleiner, B. 1995. New developments in computer software. Industrial Management & Data Systems. 95(6): 22-26.

Vasilash, G. 1995. Supply chains get smart. Production. 107(12): 48-49.

Ying, T. 1995. Computerworld. June 5, 1995. 29(33): 81.

1