Issues in pattern mining and their resolutions
C3S2E '09 Proceedings of the 2nd Canadian Conference on Computer Science and Software Engineering
Knowledge discovery from imbalanced and noisy data
Data & Knowledge Engineering
Hi-index | 0.00 |
Reliability is a key issue in knowledge discovery. However, this topic is not fully explored in data mining community. This paper takes a process perspective towards the reliability of knowledge discovery, and the reliability of extracted knowledge is evaluated by the reliability of whole knowledge discovery process. To describe the relationship between the final reliability and the reliability in each stage of process, a reliability model for generic knowledge process is proposed, and is further extended to the context of CRoss-Industry Standard Process for Data Mining (CRISP-DM). Moreover, eight factors contributing to knowledge discovery reliability are presented in the order of six phases in CRISP-DM. Based on these factors, ten suggestions on how to enhance reliability throughout knowledge discovery process are provided.