AdaptiveNormalizer2: One-pass adaptive min-max normalizalizer

In alaineiturria/otsad: Online Time Series Anomaly Detectors

One-pass adaptive min-max normalizalizer

Description

R6 class that implements our adaptation to one-pass online processing of the adaptive normalization method proposed by Gupta and Hewett. A sliding window of past w data values are maintained and the percentage change in mean values between the current and previous window is used to determine if the new or old min-max values should be used to normalize the current window. If the percentage change value exceeds the threshold, then old min-max values are replaced by the new ones.

Methods

Public methods

• AdaptiveNormalizer2$new()

• AdaptiveNormalizer2$normalize()

• AdaptiveNormalizer2$denormalize()

---

Method new()

Create a new AdaptiveNormalizalizer object.

Usage

AdaptiveNormalizer2$new(wl, threshold = 0.25, returnPoint = F)

Arguments

wl

Length of the window with historical data values.

threshold

Dissimilarity factor between the previous and current windows mean to calculate new min and max values.

returnPoint

If FALSE then normalized current window is returned, else, the normalized current data point is returned.

Returns

A new 'AdaptiveNormalizer2' object.

---

Method normalize()

Normalizes the current data value.

Usage

AdaptiveNormalizer2$normalize(x)

Arguments

x

Current data value to be normalized.

Returns

If returnPoint = FALSE then normalized current window is returned, else, normalized current data point is returned.

---

Method denormalize()

Denormalizes the current data value.

Usage

AdaptiveNormalizer2$denormalize(y)

Arguments

y

Current data value to be denormalized.

Returns

Denormalized current data point.

References

V. Gupta, R. Hewett, Adaptive Normalization in Streaming Data, in:Proceedings of the 2019 3rd International Conference on Big Data Re-search, ACM, New York, NY, USA, 2019, pp. 12-17.

Examples

normalizer <- AdaptiveNormalizer2$new(wl = 3)

normalizer$normalize(10)
normalizer$normalize(15)
normalizer$normalize(20)
normalizer$normalize(10)
normalizer$normalize(30)
normalizer$normalize(15)

normalizer$denormalize(0.25)

alaineiturria/otsad documentation built on Jan. 12, 2023, 12:26 p.m.