WaveletOutlierDetection: WaveletOutlierDetection

In Mthrun/TSAT: Time Series Analysis Tools (TSAT)

WaveletOutlierDetection

Description

Outlier detection based on a wavelet filter for signals. First, the signal is filtered. Second, the difference between signal and filtered signal is measured. If the original signal lies above the filtered signal by a factor of 1.5 (Default), it is recognized as outlier. Technical details: detection of outliers is pointwise. Each point of the signal is checked against the filtered signal. This can be done in two ways: 1. compare data point against moving average (Default) 2. compare data point against data point of filtered signal.

Usage

WaveletOutlierDetection(Data, Factor = 1.5, Strategy="mean",
Order = 7, Filter="haar", NumLevels=2, Boundary="periodic", Fast=T, PlotIt=F,
Threshold="zero", Lambda=0.05, FilterLevels="all")

Arguments

Data	Signal filtered with wavelets
Factor	Numeric defining the factor by which a data point needs to exceed the data in order to be recognized as outlier.
Strategy	Character defining the strategy by which single data points compared to data. Either, a data point is compared to the mean of the filtered data over a certain time range (moving average which time interval is defined by argument Order) around the data point or data points are compared to the filtered data points at their respective time.
Order	Integer defining the window size computed for the moving average.
Filter	Character containing name of wavelet
NumLevels	Integer determining the number of wavelet levels to create.
Boundary	A character string indicating which boundary method to use. boundary = "periodic" and boundary = "reflection" are the only supported methods at this time.
Fast	A logical flag which, if true, indicates that the pyramid algorithm is computed with an internal C function. Otherwise, only R code is used in all computations.
PlotIt	Boolean flag. If TRUE then plots will be generated. Default=F.
Threshold	Character with filtering strategy. Default: Threshold="zero". Threshold="zero": sets complete level to zero. Threshold="hard": performs hard thresholding. Threshold="soft": performs soft thresholding.
Lambda	Numeric value. If filtering uses hard or soft thresholding, Lambda indicates the p-th percentile which is used as threshold. Every value in level below this is zero (both for hard and soft thresholding). Values greater or equal are treated accordingly to soft/hard thresholding.
FilterLevels	Generic: FilterLevels="all" character: chooses all wavelet levels for filtering FilterLevels[1:n] numeric vector: chooses all wavelet levels which are given in vector.

Value

FilteredData[1:n] Signal filtered with wavelets

Author(s)

Q. Stier

References

https://cran.r-project.org/web/packages/wavelets/index.html

Examples

load("TempMelbourneAustralia")
VarO1 = TempMelbourneAustralia$Temp
VarI1 = forecast::na.interp(VarO1)
Outliers = WaveletOutlierDetection(VarI1, Order = 7)
#plot(Outliers, col = "red", ylim = c(0,10))
#lines(VarI1, col = "green")

Mthrun/TSAT documentation built on Feb. 5, 2024, 11:15 p.m.