Authors:
Lauri Tuovinen
;
Juha Röning
and
Perttu Laurinen
Affiliation:
University of Oulu, Finland
Keyword(s):
Data mining, Application framework, Pipes and filters, Tool integration, Proxy pattern.
Related
Ontology
Subjects/Areas/Topics:
Business Analytics
;
Communication and Software Technologies and Architectures
;
Data Engineering
;
Data Exchange and Integration
;
Data Warehouses and Data Mining
;
e-Business
;
Embedded Communications Systems
;
Enterprise Information Systems
;
Software Architectures
;
Telecommunications
Abstract:
An important phase in the development of a data mining application is algorithm selection: for any given data mining task there is likely to be a range of different model types available, as well as a number of different methods for constructing the models. Choosing the one that best accomplishes the task is not trivial and generally involves trials and comparisons of different configurations. It is often convenient to perform the trials on a platform other than the ultimate implementation technology of the application; for example, the application may be implemented in a general-purpose programming language such as C++ while model prototyping is carried out in a scientific computing environment such as MATLAB. Smart Archive, a data mining application framework for Java and C++, allows MATLAB and other external tools to be integrated into applications via special proxies known as gateway filters. To the framework the gateway filters appear no different from algorithms implemented on
the framework platform, so it is possible to build a full application prototype early on and then, once the algorithms to be used have been selected, to turn the prototype into a deployable application simply by replacing proxies with natively implemented filters. Thus the framework comprehensively supports the various steps of application programming, from algorithm design and prototype building to final implementation.
(More)