Workshop at the Fourth International Conference on Knowledge Discovery
and Data Mining
Sponsored by the Association for the Advancement of Artificial Intelligence
KDD1998 Workshops were held on Monday, August 31, 1998 in New York, New York. The workshops that were held are listed below:
- Data Mining in Finance (KW1)
Organizer: Tae Horn Hann
- Distributed Data Mining (KW2)
Organizers: Hillol Kargupta and Philip Chan
- Keys to the Commercial Success of Data Mining (KW3)
Organizers: Kurt Thearling and Roger M. Stein
Data Mining in Finance
In recent years data mining is being increasingly applied in finance, especially to support financial asset management and risk management. It is considered by many financial management institutions as an innovative technology to support conventional quantitative techniques. Its use in computational finance will have a major impact in the modeling of currency markets, in tactical asset allocation, bond and stock valuation and portfolio optimization. In addition, the application of data mining for scoring tasks delivers valuable support for the management of client credit risk and fraud detection.
On the other hand, there are interesting debates in finance that have caused controversial discussion over time and can significantly affect the application of data mining in finance. To these debates belong: 1) Efficiency of financial markets. According to the information efficiency of financial markets, price changes are unpredictable. If this is the case, any effort putting on mining historical data which aims to predict the future development is useless. 2) Nonlinearity of financial markets. If price movements in financial markets are linear, then the application of complicated data mining algorithms like neural networks is not necessary at all. Simpler methods like linear regression would be sufficient.
The goal of this workshop is to provide an informal forum for researchers and practitioners to discuss the following topics:
- Are there any special aspects of data mining in finance which are not typical for data mining in other fields (for examples technical fields, health, etc.)?
- What are the characteristics of the successful applications of data mining in finance?
- What are the typical pitfalls?
- Are the financial markets efficient? What are the pro and contra arguments?
- Are the financial markets nonlinear? Is there any positive or negative evidence about this debate?
This workshop addresses practitioners as well as researchers from those communities which contribute to this topic such as finance, econometrics, statistics and information systems.
The workshop includes invited talks, contributed talks, two panel discussions and a final discussion. The invited and contributed talks discuss lessons learned from realized data mining projects in finance, success stories, pitfalls, special aspects of data mining in finance, task complexity, dynamic aspects and legal issues. In two panel discussions the controversial point of view about efficiency and non-linearity of finance markets are discussed by eight panelists.
Distributed Data Mining
Automated detection of patterns from large amount of data is often called data mining. As computing and communication are increasingly converging to each other, mining data, stored in distributed databases with adequate attention to security related issues, is of growing interest. Distributed data mining (DDM) systems are finding an increasing number of applications in popular Intranet/Internet environments, data-mart based warehousing architectures, network intrusion detection, geographical information systems and many others. This workshop will provide a platform for discussing theoretical and applied research issues in DDM. The topics of interest include, but are not limited to:
- Theory and foundation issues in DDM: Problem decomposability and data distribution; complexity issues in DDM; representational issues.
- Methods and algorithms: Distributed algorithms for popular data mining techniques (e.g. association rules, classifiers, clustering); techniques for communication minimization, cooperative learning.
- Software agents and DDM: Agent based approaches in DDM. Agent interaction: cooperation, collaboration, negotiation, organizational behavior.
- DDM for spatial data: DDM in geographical information databases.
- Architectural issues in DDM: Architecture, control, security, communication issues.
- Experimental DDM systems: Large experimental systems, performance, design issues.
- Applications of DDM: Application of DDM in business, science, engineering, and medicine.
- Human interaction in DDM: Human-DDM interface, multi-user interaction in DDM.
- Distributed data mining on the Internet.
- Parallel Data mining: Parallel data mining algorithms, applications; high performance computing in DDM
Keys to the Commercial Success of Data Mining
Successful data mining in business no longer comes down to having a hot algorithm and an experienced modeler. Business users care about things such as database support, application integration, business templates, flexibility, scalability, profitability, and other issues not historically concerning the KDD community.
A number of the issues that we hope will get addressed at the workshop are described in Kurt Thearling's article "Some Thoughts on the Current State of Data Mining Software Applications" and in an interview given by Roger Stein.
The goal is to bring together a diverse group of forty to fifty developers, users, and integrators of business data mining applications. The workshop will consist of in-depth case studies and analyses, invited speakers, and panel sessions. Time will also be set aside for discussions.
Approximately half of the participants will come from the data mining development community with the other half coming from the data mining business user community. Business users attending will be selected from diverse industries such as banking, retail, insurance, government, internet services, telecom, etc.
Workshop Technical Reports
Some KDD workshops are available as technical reports. For contents and ordering information, consult the AAAI Press catalog.