|
| | | |
Adaptive Spike Detection for Resilient Data Stream Mining
Phua, C., Smith-Miles, K., Lee, V.C.S. and Gayler, R.
Automated adversarial detection systems can fail
when under attack by adversaries. As part of a resilient
data stream mining system to reduce the possibility
of such failure, adaptive spike detection is
attribute ranking and selection without class-labels.
The first part of adaptive spike detection requires
weighing all attributes for spiky-ness to rank them.
The second part involves filtering some attributes
with extreme weights to choose the best ones for computing
each example's suspicion score. Within an
identity crime detection domain, adaptive spike detection
is validated on a few million real credit applications
with adversarial activity. The results are
F-measure curves on eleven experiments and relative
weights discussion on the best experiment. The results
reinforce adaptive spike detection's effectiveness
for class-label-free attribute ranking and selection. |
Cite as: Phua, C., Smith-Miles, K., Lee, V.C.S. and Gayler, R. (2007). Adaptive Spike Detection for Resilient Data Stream Mining. In Proc. Sixth Australasian Data Mining Conference (AusDM 2007), Gold Coast, Australia. CRPIT, 70. Christen, P., Kennedy, P. J., Li, J., Kolyshkina, I. and Williams, G. J., Eds. ACS. 181-188. |
(from crpit.com)
(local if available)
|
|