forestFloor (version 1.5)

plot.forestFloor: Metod: plot.forestFloor

Description

Method to plot an object of forestFloor-class. Plot partial feature contributions of the most important variables. Colour gradients can be applied two show possible interactions.

Usage

## S3 method for class 'forestFloor':
plot(x,
                            plot_seq=NULL, 
                            limitY=TRUE,
                            order_by_importance=TRUE, 
                            cropXaxes=NULL, 
                            crop_limit=4,
                            plot_GOF = FALSE,
                            GOF_col = "#33333399",
                            ...)

Arguments

x

foretFloor-object, also abbrivated ff.. Computed topology of randomForest-model, the output from the forestFloor function includes also X and Y and importance data

plot_seq

a numeric vector describing which variables and in what sequence to plot, ordered by importance as default, order_by_importance = F then by feature/coloumn order of training data.

limitY

TRUE/FLASE, constrain all Yaxis to same limits to ensure relevance of low importance features is not overinterpreted

order_by_importance

TRUE / FALSE should plotting and plot_seq be ordered after importance. Most important feature plot first(TRUE)

cropXaxes

a vector of indice numbers of which zooming of x.axis should look away from outliers

crop_limit

a number often between 1.5 and 5, referring limit in std.devs from the mean defining outliers if limit = 2, above selected plots will zoom to +/- 2 std.dev of the respective features.

plot_GOF

Booleen TRUE/FALSE. Should the goodness of fit be plotted as a line?

GOF_col

Color of plotted GOF line

...

... other arguments passed to generic plot functions

Details

The method plot.forestFloor visualizes partial plots of the most important variables first. Partial dependence plots are available in the randomForest package. But such plots are single lines(1d-slices) and do not answer the question: Is this partial function(PF) a fair generalization or subject to global or local interactions.

Examples

#simulate data
obs=1000
vars = 6 
X = data.frame(replicate(vars,rnorm(obs)))
Y = with(X, X1^2 + sin(X2*pi) + 2 * X3 * X4 + 0.5 * rnorm(obs))

#grow a forest, remeber to include inbag
rfo=randomForest::randomForest(X,Y,keep.inbag=TRUE)

#compute topology
ff = forestFloor(rfo,X)

#print forestFloor
print(ff) 

#plot partial functions of most important variables first
plot(ff,order_by_importance=TRUE) 

#Non interacting functions are well displayed, whereas X3 and X4 are not
#by applying different colourgradient, interactions reveal themself 
#also a k-nearest neighbor fit is applied to evaluate goodness of fit
Col=fcol(ff,3,orderByImportance=FALSE)
plot(ff,col=Col,plot_GOF=TRUE) 

#if needed, k-nearest neighbor parameters for goodness-of-fit can be access through convolute_ff
#a new fit will be calculated and added to forstFloor object as ff$FCfit
ff = convolute_ff(ff,userArgs.kknn=alist(kernel="epanechnikov",kmax=5))
plot(ff,col=Col,plot_GOF=TRUE)