teda_r: Create teda recursive object from observation (+ state)

In NERC-CEH/teda-r: An Implementation of the Typicality and Eccentricity Data Analysis Framework

Description

A recursive method that takes the state variables of previous mean, previous variance, and the current timestep position, along with the current observation. It returns a teda recursive object. Currently only a univariate implementation.

Usage

teda_r(curr_observation, previous_mean = curr_observation, previous_var = 0,
  k = 1, dist_type = "Euclidean")

Arguments

curr_observation	A single observation, the most recent in a series
previous_mean	The mean value returned by the previous call to this function, if no previous calls, default value is used.
previous_var	The variance value returned by the previous call to this function, if no previous calls, default value is used.
k	The count of observations processed by the recursive function, including the current observation
dist_type	A string representing the distance metric to use, default value (and currently only supported value) is "Euclidean"

Details

The function has two intended ways of use: on the first pass, it only takes the observation value as a paramter and the rest are provided by defaults, on all other passes, it takes the current observation, the previous mean and variance values, and the current k (number of observations) which includes the current observation.

On return, the teda recursive object holds:

• the current observation

• the current mean

• the current variance

• the current observation's eccentricity

• the current observation's typicality

• the current observation's normalised eccentricity

• the current observation's normalised typicality

• whether the current observation is an outlier

• the current outlier threshold

• the next timestep value, k+1

It provides generic functions for print and summary, at this moment both provide the same outout.

Value

The teda recursive object

References

Bezerra, C.G., Costa, B.S.J., Guedes, L.A. and Angelov, P.P., 2016, May. A new evolving clustering algorithm for online data streams. In Evolving and Adaptive Intelligent Systems (EAIS), 2016 IEEE Conference on (pp. 162-168). IEEE. DOI: 10.1109/EAIS.2016.7502508

See Also

Other TEDA.functions: teda_b

Examples

vec = c(20, 12, 10, 20)
a = teda_r(vec[1])
b = teda_r(vec[2],
           a$curr_mean,
           a$curr_var,
           a$next_k)
c = teda_r(vec[3],
           b$curr_mean,
           b$curr_var,
           b$next_k)

d = teda_r(vec[4],
           c$curr_mean,
           c$curr_var,
           c$next_k)
summary(d)

NERC-CEH/teda-r documentation built on May 7, 2019, 6:01 p.m.