iter_pivot: Iterative algorithm using pivot features

Description Usage Arguments Value Author(s) Examples

View source: R/iter_pivot.R

Description

An iterative semi-supervised learning approach using pivot features. Uses a random forest classifier to train and predict probabilities.

Usage

1
iter_pivot(source, target, x_pivot, x_nonpivot, rho = 3)

Arguments

source

A data frame containing observations and features for the source domains. Must contain a column named "Domain" which specifies the domain each observation belongs to and "Class" which gives the response.

target

A data frame containing unlabelled observations and features for the target domain.

x_pivot

A vector of strings. Gives the column names of features with common distribution between source and target domains.

x_nonpivot

A vector of strings. Gives the column names of features that have a shifted distribution between the target and source domains.

rho

Number of observations to be added to labelled target training set (actual number added is 4*rho).

Value

iter_pivot returns an object of class "iter_pivot".

The function plot produces a plot showing predictions for the most recent iteration in the two feature spaces.

An object of class "iter_pivot" is a list containing the following components:

Author(s)

Cameron Roach

Examples

1
2
3
4
5
6
source <- sim_pivot_data[sim_pivot_data$Domain == "Source",]
target <- sim_pivot_data[sim_pivot_data$Domain == "Target",]
predictions <- iter_pivot(source = source, target = target,
                          x_pivot = "Feature1", x_nonpivot = "Feature2",
                          rho = 3)
#TODO: FIX plot(predictions, aes(x = Feature1, y = Feature2))

camroach87/semisupervisr documentation built on June 4, 2017, 8:55 a.m.