Mining High Impact Exceptional Behavior Patterns
In the real world, exceptional behavior can be seen in many situations such as security-oriented fields. Such behavior is rare and dispersed, while some of them may be associated with significant impact on the society. A typical example is the event September 11. The key feature of the above rare but significant behavior is its high potential to be linked with some significant impact. Identifying such particular behavior before generating impact on the world is very important. In this paper, we develop several types of high impact exceptional behavior patterns. The patterns include frequent behavior patterns which are associated with either positive or negative impact, and frequent behavior patterns that lead to both positive and negative impact. Our experiments in mining debt-associated customer behavior in social-security areas show the above approaches are useful in identifying exceptional behavior to deeply understand customer behavior and streamline business process.
KeywordsFrequent Itemsets Relative Risk Ratio Slide Time Window Exceptional Behavior Window Mode
Unable to display preview. Download preview PDF.
- 1.Zhao, Y., Cao, L.: Full report: Improving income reporting (May 31, 2006)Google Scholar
- 2.Cao, L., Zhang, C.: Domain-driven data mining: a practical methodology. Int. J. of Data Warehousing and Mining (2006)Google Scholar
- 3.Centrelink. Integrated activity management developer guide, Technical Report, 30 (September 1999)Google Scholar
- 4.Centrelink: Summary report: Improving income reporting (June 2006)Google Scholar
- 7.Zhang, J., Bloedorn, E., Rosen, L., Venese, D.:  Zhang, J. In: Perner, P. (ed.) ICDM 2004. LNCS (LNAI), vol. 3275, pp. 571–574. Springer, Heidelberg (2004)Google Scholar
- 8.Cao, L., Zhao, Y., Zhang, C., Zhang, H.: Activity Mining: from Activities to Actions. International Journal of Information Technology & Decision Making 7(2) (2008)Google Scholar
- 9.Cao, L., Zhao, Y., Zhang, C.: Mining Impact-Targeted Activity Patterns in Imbalanced Data. IEEE Trans. on Knowledge and Data Engineering (to appear)Google Scholar