ContextualAnomalyDetector function

Contextual Anomaly Detector - Open Source (CAD)

ContextualAnomalyDetector calculates the anomaly score of a dataset using the notion of contexts conformed by facts and provides probabilistic abnormality scores.

ContextualAnomalyDetector(data, rest.period = max(min(150,
  round(length(data) * 0.03), 1)), max.left.semicontexts = 7,
  max.active.neurons = 15, num.norm.value.bits = 3,
  base.threshold = 0.75, min.value = min(data, na.rm = T),
  max.value = max(data, na.rm = T), python.object = NULL, lib = 0)

Arguments

• data: Numerical vector with training and test dataset.
• rest.period: Training period after an anomaly.
• max.left.semicontexts: Number of semicontexts that should be maintained in memory.
• max.active.neurons: Number of neurons of the model.
• num.norm.value.bits: Granularity of the transformation into discrete values
• base.threshold: Threshold to be considered an anomaly.
• min.value: Minimum expected value.
• max.value: Maximum expected value.
• python.object: Python object for incremental processing.
• lib: 0 to run the original python script, 1 to get the same results on all operating systems.

Returns

List - result: Data frame with anomaly.score and is.anomaly comparing the anomaly score with base.threshold.

• python.object: ContextualAnomalyDetector Python object used in online anomaly detection

Details

data must be a numerical vector without NA values. threshold must be a numeric value between 0 and 1. If the anomaly score obtained for an observation is greater than the threshold, the observation will be considered abnormal. Requires hashlib (included in python installation) and bencode-python3 (which can be installed using pip) python libraries.

Examples

## Generate data
set.seed(100)
n <- 200
x <- sample(1:100, n, replace = TRUE)
x[70:90] <- sample(110:115, 21, replace = TRUE)
x[25] <- 200
x[150] <- 170
df <- data.frame(timestamp = 1:n, value = x)

## Calculate anomalies

   result <- ContextualAnomalyDetector(data = df$value, rest.period = 10, base.threshold = 0.9)

   ## Plot results
   res <- cbind(df, result$result)
   PlotDetections(res, title = "CAD_OSE ANOMALY DETECTOR")

References

Smirnov, M. (2018). CAD: Contextual Anomaly Detector. https://github.com/smirmik/CAD