streamMOA (version 1.2-3)

DSC_MCOD_MOA: Micro-cluster Continuous Outlier Detector (MCOD)

Description

Class interfaces the MOA implementation of the MCOD algorithm for distance-based data stream outlier detection.

Usage

DSC_MCOD(r = 0.1, t = 50, w = 1000, recheck_outliers = TRUE)

Arguments

r

Defines the micro-cluster radius

t

Defines the number of neighbors (k in the article)

w

Defines the window width in data points

recheck_outliers

Defines that the MCOD algorithm allows re-checking of detected outliers.

Value

An object of class DSC_MCOD (subclass of DSC_SinglePass, DSC_Outlier, DSC_Micro, DSC_MOA and DSC).

Details

The algorithm detects density-based outliers. An object \(x\) is defined to be an outlier if there are less than \(t\) objects lying at distance at most \(r\) from \(x\).

References

Kontaki M, Gounaris A, Papadopoulos AN, Tsichlas K, and Manolopoulos Y (2016). "Efficient and flexible algorithms for monitoring distance-based outliers over data streams." In "Information systems", vol. 55, pp. 37-53.10.1109/ICDE.2011.5767923

See Also

DSC, DSC_Micro, DSC_MOA, DSC_SinglePass, DSC_Outlier

Examples

# NOT RUN {
# two-stage example
stream <- DSD_Gaussians(k = 3, d = 2,
            separation_type = "Mahalanobis", separation = 4,
            space_limit = c(0, 30), variance_limit = 0.8,
            outliers = 10,
            outlier_options = list(outlier_horizon = 1000))

mic_c <- DSC_MCOD(r = 1, t = 10, w = 1000)
mac_c <- DSC_Kmeans(3)
c <- DSC_TwoStage(mic_c, mac_c)

evaluate(c, stream, n = 1000, type = "macro",
  measure = c("crand","outlierjaccard"))

reset_stream(stream)
plot(c, stream, n = 1000, type = "all")
# }