User-Driven Ranking for Measuring the Interestingness of Knowledge Patterns

Choosing thresholds for knowledge discovery algorithms to achieve a set of results which either solve a specific problem or are optimised towards the users desires requires the re-run of the algorithm, which is a tedious and time-consuming procedure. Within this paper an importance based interestingness measure is introduced that can be applied as a posterior phase. In addition three manipulative operations (balancing, boosting and inversion) are outlined operating upon results of data mining exercises to allow the interpretation of the same result from different viewpoints. The overall framework represents a novel user-driven ranking procedure, allowing measuring the interestingness of knowledge patterns utilising a combination of qualitative and quantitative indicators as well as a user driven importance or interestingness value.

[1]  Myra Spiliopoulou,et al.  Web Usage Analysis and User Profiling , 2002, Lecture Notes in Computer Science.

[2]  Myra Spiliopoulou,et al.  WUM: A tool for Web Utilization analysis , 1999 .

[3]  Abraham Silberschatz,et al.  What Makes Patterns Interesting in Knowledge Discovery Systems , 1996, IEEE Trans. Knowl. Data Eng..

[4]  Nong Ye,et al.  The Handbook of Data Mining , 2003 .

[5]  Ramakrishnan Srikant,et al.  Mining sequential patterns , 1995, Proceedings of the Eleventh International Conference on Data Engineering.

[6]  Maurice Mulvenna,et al.  Navigation Pattern Discovery from Internet Data , 1999 .

[7]  William Frawley,et al.  Knowledge Discovery in Databases , 1991 .

[8]  Jörg Rech,et al.  Knowledge Discovery in Databases , 2001, Künstliche Intell..

[9]  Philip S. Yu,et al.  Efficient mining of weighted association rules (WAR) , 2000, KDD '00.