caret (version 6.0-70)

lift: Lift Plot

Description

For classification models, this function creates a 'lift plot' that describes how well a model ranks samples for one class

Usage

lift(x, ...)
"lift"(x, data = NULL, class = NULL, subset = TRUE, lattice.options = NULL,  cuts = NULL, labels = NULL,  ...)
"xyplot"(x, data, plot = "gain", values = NULL, ...)

Arguments

x

a lattice formula (see xyplot for syntax) where the left-hand side of the formula is a factor class variable of the observed outcome and the right-hand side specifies one or model columns corresponding to a numeric ranking variable for a model (e.g. class probabilities). The classification variable should have two levels.

data

For lift.formula, a data frame (or more precisely, anything that is a valid envir argument in eval, e.g., a list or an environment) containing values for any variables in the formula, as well as groups and subset if applicable. If not found in data, or if data is unspecified, the variables are looked for in the environment of the formula. This argument is not used for xyplot.lift.

class

a character string for the class of interest

subset

An expression that evaluates to a logical or integer indexing vector. It is evaluated in data. Only the resulting rows of data are used for the plot.

lattice.options

A list that could be supplied to lattice.options

cuts

If a single value is given, a sequence of values between 0 and 1 are created with length cuts. If a vector, these values are used as the cuts. If NULL, each unique value of the model prediction is used. This is helpful when the data set is large.

labels

A named list of labels for keys. The list should have an element for each term on the right-hand side of the formula and the names should match the names of the models.

plot

Either "gain" (the default) or "lift". The former plots the number of samples called events versus the event rate while the latter shows the event cut-off versus the lift statistic.

values

A vector of numbers between 0 and 100 specifying reference values for the percentage of samples found (i.e. the y-axis). Corresponding points on the x-axis are found via interpolation and line segments are shown to indicate how many samples must be tested before these percentages are found. The lines use either the plot.line or superpose.line component of the current lattice theme to draw the lines (depending on whether groups were used. These values are only used when type = "gain".

...

options to pass through to xyplot or the panel function (not used in lift.formula).

Value

lift.formula returns a list with elements:

data

the data used for plotting

cuts

the number of cuts

class

the event class

probNames

the names of the model probabilities

pct

the baseline event rate

xyplot.lift returns a lattice object

Details

lift.formula is used to process the data and xyplot.lift is used to create the plot.

To construct data for the the lift and gain plots, the following steps are used for each model:

1. The data are ordered by the numeric model prediction used on the right-hand side of the model formula
2. Each unique value of the score is treated as a cut point
3. The number of samples with true results equal to class are determined
4. The lift is calculated as the ratio of the percentage of samples in each split corresponding to class over the same percentage in the entire data set

lift with plot = "gain" produces a plot of the cumulative lift values by the percentage of samples evaluated while plot = "lift" shows the cut point value versus the lift statistic.

This implementation uses the lattice function xyplot, so plot elements can be changed via panel functions, trellis.par.set or other means. lift uses the panel function panel.lift2 by default, but it can be changes using update.trellis (see the examples in panel.lift2).

The following elements are set by default in the plot but can be changed by passing new values into xyplot.lift: xlab = "% Samples Tested", ylab = "% Samples Found", type = "S", ylim = extendrange(c(0, 100)) and xlim = extendrange(c(0, 100)).

See Also

xyplot, trellis.par.set

Examples

set.seed(1)
simulated <- data.frame(obs = factor(rep(letters[1:2], each = 100)),
                        perfect = sort(runif(200), decreasing = TRUE),
                        random = runif(200))

lift1 <- lift(obs ~ random, data = simulated)
lift1
xyplot(lift1)

lift2 <- lift(obs ~ random + perfect, data = simulated)
lift2
xyplot(lift2, auto.key = list(columns = 2))

xyplot(lift2, auto.key = list(columns = 2), value = c(10, 30))

xyplot(lift2, plot = "lift", auto.key = list(columns = 2))