rminer (version 1.1)

Importance: Measure input importance given a supervised data mining model.

Description

Measure input importance given a supervised data mining model.

Usage

Importance(M, data, RealL = 6, method = "sens", measure = "gradient", 
           sampling = "regular", baseline = "mean", responses = TRUE, 
           outindex = NULL, task = "default", PRED = NULL, 
           interactions = NULL)

Arguments

M

fitted model, typically is the object returned by fit. Can also be any fitted model (i.e. not from rminer), provided that the predict function PRED is defined (see examples for details).

data

training data (the same data.frame that was used to fit the model, currently only used to add data histogram to VEC curve).

RealL

for numeric inputs, the number of sensitivity analysis levels (e.g. 6). Note: you need to use RealL>=2.

method

input importance method. Options are:

• sens -- sensitivity analysis
• sensv -- equal tosensbut setsmeasure="variance".
• sensg -- equal tosensbut setsmeasure="gradient"

measure

sensitivity analysis measure (used to measure input importance). Options are:

• gradient -- average absolute gradient (y_i+1-y_i) of the responses.
• variance -- variance of the responses.
• range -- maximum - minimum of

sampling

for numeric inputs, the sampling scan function. Options are:

• regular -- regular sequence (uniform distribution).
• quantile -- sample values from the input that are more closer to the variable distribution indata

baseline

baseline vector used during the sensitivity analysis. Options are:

• mean -- uses a vector with the mean values of each attribute fromdata.
• median -- uses a vector with the median values of each attribute from<

responses

if TRUE then all sensitivity analysis responses are stored and returned.

outindex

the output index (column) of data if M is not a model object (returned by fit).

task

the task as defined in fit if M is not a model object (returned by fit).

PRED

the prediction function of M, if M is not a model object (returned by fit). Note: this function should behave like the rminer predict-methods, i.e. return a

interactions

numeric vector with the attributes (columns) used by Ith-D sensitivity analysis (2-D or higher):

• ifNULLthen only a 1-D sensitivity analysis is performed.
• iflength(interactions)==1?then a "special" 2-D sensitivity

Value

• A list with the components:
  • $value -- numeric vector with the computed sensitivity analysis measure for each attribute.
  • $imp -- numeric vector with the relative importance for each attribute.
  • $sresponses -- vector list as described in the Value documentation ofmining.

Details

This function provides several algorithms for measuring input importance of supervised data mining models. A particular emphasis is given on sensitivity analysis (SA), which is a simple method that measures the effects on the output of a given model when the inputs are varied through their range of values. Check the reference for more details.

References

• To cite the Importance function or sensitivity analysis method, please use: P. Cortez and M. Embrechts. Opening Black Box Data Mining Models Using Sensitivity Analysis. In Proceedings of the 2011 IEEE Symposium on Computational Intelligence and Data Mining (CIDM), pp. 341-348, Paris, France, April, 2011. http://www3.dsi.uminho.pt/pcortez/

See Also

vecplot, fit, mining, mgraph, mmetric, savemining.

Examples

### Typical use under rminer:
# 1st example, regression, 1-D sensitivity analysis
data(sin1reg) # x1 should account for 70%, x2 for 30% and x3 for 0%.
M=fit(y~.,sin1reg,model="svm")
I=Importance(M,sin1reg,method="sens",measure="gradient") # 1-D SA
print(I)
L=list(runs=1,sen=t(I$imp),sresponses=I$sresponses)
mgraph(L,graph="IMP",leg=names(sin1reg),col="gray",Grid=10)
mgraph(L,graph="VEC",xval=1,Grid=10,data=sin1reg) # or:
vecplot(I,xval=1,Grid=10,data=sin1reg,datacol="gray") # the same graph
vecplot(I,xval=c(1,2,3),pch=c(1,2,3),Grid=10,
leg=list(pos="bottomright",leg=c("x1","x2","x3"))) # all x1, x2 and x3 VEC curves

# 2nd example, regression, 2-D sensitivity analysis with 
# the most relevant input (x1, index 1):
I2=Importance(M,sin1reg,method="sensg",interactions=which.max(I$imp))
print(I2)
# influence of x1 and x2 over y
vecplot(I2,graph="VEC3",xval=2) # VEC surface
vecplot(I2,graph="VECC",xval=2) # VEC contour
# influence of x1 and x3 over y (influence of x3 is small random noise, 0%):
vecplot(I2,graph="VEC3",xval=3)
vecplot(I2,graph="VECC",xval=3)

# 3rd example, regression, full 3-D sensitivity analysis
I3=Importance(M,sin1reg,method="sensg",interactions=c(1,2,3))
print(I3)
I3_1d=avg_imp(I3,c(1)) # 1-D averaging under x1
vecplot(I3_1d,graph="VEC",xval=1,Grid=10)
I3_2d=avg_imp(I3,c(1,2)) # 2-D averaging under the pair x1,x2
vecplot(I3_2d,graph="VEC3")

### If you want to use Importance over your own model:
# 1st example, regression, uses the theoretical sin1reg function
mypred=function(M,data)
{ return (M[1]*sin(pi*data[,1]/M[3])+M[2]*sin(pi*data[,2]/M[3])) }
M=c(0.7,0.3,2000)
# 4 is the column index of y
I=Importance(M,sin1reg,method="sens",measure="gradient",PRED=mypred,outindex=4) 
print(I$imp) # x1=72.3% and x2=27.7%
L=list(runs=1,sen=t(I$imp),sresponses=I$sresponses)
mgraph(L,graph="IMP",leg=names(sin1reg),col="gray",Grid=10)
mgraph(L,graph="VEC",xval=1,Grid=10) # equal to:
vecplot(I,graph="VEC",xval=1,Grid=10)

# 2nd example, 3-class classification for iris and lda model:
data(iris)
library(MASS)
predlda=function(M,data) # the PRED function
{ return (predict(M,data)$posterior) }
LDA=lda(Species ~ .,iris, prior = c(1,1,1)/3)
# 4 is the column index of Species
I=Importance(LDA,iris,method="sensg",PRED=predlda,outindex=4)
vecplot(I,graph="VEC",xval=1,Grid=10,TC=1,
main="1-D VEC for Sepal.Lenght (x-axis) influence in setosa (prob.)")

# 3rd example, binary classification for setosa iris and lda model:
iris2=iris;iris2$Species=factor(iris$Species=="setosa")
predlda2=function(M,data) # the PRED function
{ return (predict(M,data)$class) }
LDA2=lda(Species ~ .,iris2)
I=Importance(LDA2,iris2,method="sensg",PRED=predlda2,outindex=4)
vecplot(I,graph="VEC",xval=1,
main="1-D VEC for Sepal.Lenght (x-axis) influence in setosa (class)",Grid=10)