Data mining applications for diverse industrial application domains with smart archive
Lauri Tuovinen, Perttu Laurinen, Ilmari Juutilainen, Juha Röning · 2008
The increasing significance of data mining in many appli-cation domains of computational technology has resulted in a considerable body of work concerned with design-ing architectural models and collections of reusable soft-ware components to allow for more rapid deployment of data mining methods wherever they are deemed useful. Early work involved integrating machine learning algo-rithms and knowledge management features into class li-braries; more recently the data mining research community has progressed towards application frameworks, which are based on the notion of a reusable high-level design rather than specific algorithms. Smart Archive provides such a design, fine-tuned for applications that process continuous measurements and make use of historical data. In this pa-per a pre-existing foundational framework is extended with a new layer of services and the overall system architecture established by the framework is examined. Two case stud-ies drawn from diverse application domains—steelmaking and personal fitness products—are presented in order to validate the proposed design. The case studies show that Smart Archive is beneficial in integration, coding and test-ing tasks and suitable for both online and batch processing and for both localized and distributed applications.