Knowledge reuse in data mining projects and its practical applications

Abstract

The objective of this paper is providing an integrated environment for knowledge reuse in KDD, for preventing recurrence of known errors and reinforcing project successes, based on previous experience. It combines methodologies from project management, data warehousing, mining and knowledge representation. Different from purely algorithmic papers, this one focuses on performance metrics used for managerial such as the time taken for solution development, the amount of files not automatically managed and other, while preserving equivalent performance on the technical solution quality metrics. This environment has been validated with metadata collected from previous KDD projects developed and deployed for real world applications by the development team members. The case study carried out in actual contracted projects have shown that this environment assesses the risk of failure for new projects, controls and documents all the KDD project development process and helps understanding the conditions that lead KDD projects to success or failure. © 2009 Springer Berlin Heidelberg.