Data Mining, Validation, and Collaborative Knowledge Capture

Abstract

For large-scale data mining, utilizing data from ubiquitous and mixed-structured data sources, the extraction and integration into a comprehensive data-warehouse is usually of prime importance. Then, appropriate methods for validation and potential refinement are essential. This chapter describes an approach for integrating data mining, information extraction, and validation with collaborative knowledge management and capture in order to improve the data acquisition processes. For collaboration, a semantic wiki-enabled system for knowledge and experience management is presented. The proposed approach applies information extraction techniques together with pattern mining methods for initial data validation and is applicable for heterogeneous sources, i.e., capable of integrating structured and unstructured data. The methods are integrated into an incremental process providing for continuous validation options. The approach has been developed in a health informatics context: The results of a medical application demonstrate that pattern mining and the applied rule-based information extraction methods are well suited for discovering, extracting and validating clinically relevant knowledge, as well as the applicability of the knowledge capture approach. The chapter presents experiences using a case-study in the medical domain of sonography.