phi.control: Generation of relevance function

In nunompmoniz/IRon: Solving Imbalanced Regression Tasks

Generation of relevance function

Description

This procedure enables the generation of a relevance function that performs a mapping between the values in a given target variable and a relevance value that is bounded by 0 (minimum relevance) and 1 (maximum relevance). This may be obtained automatically (based on the distribution of the target variable) or by the user defining the relevance values of a given set of target values - the remaining values will be interpolated.

Usage

phi.control(
  y,
  phi.parms,
  method = phiMethods,
  extr.type = NULL,
  control.pts = NULL,
  asym = TRUE,
  ...
)

Arguments

y	The target variable of a given data set
phi.parms	The relevance function providing the data points where the pairs of values-relevance are known
method	The method used to generate the relevance function (extremes or range)
extr.type	Type of extremes to be considered: low, high or both (default)
control.pts	Parameter required when using 'range' method, representing a 3-column matrix of y-value, corresponding relevance value (between 0 and 1), and the derivative of such relevance value
asym	Boolean for assymetric interpolation. Default TRUE, uses adjusted boxplot. When FALSE, uses standard boxplot.
...	Misc data to be added to the relevance function

Value

A list with three slots with information concerning the relevance function

method	The method used to generate the relevance function (extremes or range)
npts	?
control.pts	Three sets of values identifying the target value-relevance-derivate for the first low extreme value, the median, and first high extreme value

Examples

library(IRon)

data(accel)

ind <- sample(1:nrow(accel),0.75*nrow(accel))

train <- accel[ind,]
test <- accel[-ind,]

ph <- phi.control(train$acceleration); phiPlot(test$acceleration, ph)
ph <- phi.control(train$acceleration, extr.type="high"); phiPlot(test$acceleration, ph)
ph <- phi.control(train$acceleration, method="range",
  control.pts=matrix(c(10,0,0,15,1,0),byrow=TRUE,ncol=3)); phiPlot(test$acceleration, ph)

nunompmoniz/IRon documentation built on April 24, 2023, 1:20 p.m.