Using caret on the Titanic data from Kaggle, I tried various models,
including rfRules which produces a model, partly described as such:
> caret.rfRules.cv$finalModel
$model
len freq err
[1,] "2" "0.0368" "0"
[2,] "2" "0.032" "0.05"
[3,] "2" "0.1824" "0.0526315789473685"
[4,] "4" "0.0656" "0.0975609756097561"
[5,] "4" "0.0304" "0.105263157894737"
[6,] "3" "0.4384" "0.105839416058394"
[7,] "3" "0.0112" "0.142857142857143"
[8,] "4" "0.0256" "0.1875"
[9,] "3" "0.1088" "0.279411764705882"
[10,] "3" "0.056" "0.342857142857143"
[11,] "1" "0.0128" "0.25"
condition pred
[1,] "X[,4]<=7 & X[,11]<=4.5"
"1"
[2,] "X[,7]<=31.33125 & X[,11]>4.5"
"0"
[3,] "X[,2]<=0.5 & X[,3]<=0.5"
"1"
[4,] "X[,2]>0.5 & X[,4]<=30.5 & X[,5]>0.5 &
X[,9]>0.5" "0"
[5,] "X[,3]<=0.5 & X[,4]<=30.2031919426199 & X[,4]>21.5
& X[,9]<=0.5" "1"
[6,] "X[,3]>0.5 & X[,4]>9.5 & X[,7]<=26.26875"
"0"
[7,] "X[,2]>0.5 & X[,7]>13.90835 & X[,7]<=15.3729"
"0"
[8,] "X[,4]<=40.8653667208804 & X[,4]>25 & X[,7]>26.14375
& X[,11]<=1.5" "1"
[9,] "X[,3]>0.5 & X[,4]>8.16718191075288 &
X[,4]<=77" "0"
[10,] "X[,3]<=0.5 & X[,4]<=38.5 & X[,4]>12.5"
"1"
[11,] "X[,1]==X[,1]"
"0"
[...]
Does that 11th row make sense? X[,1]==X[,1] will always be true, so
is that saying anything? Or is it a case of a model for prediction
being useless for inference?
> version
_
platform x86_64-pc-linux-gnu
arch x86_64
os linux-gnu
system x86_64, linux-gnu
status
major 3
minor 4.1
year 2017
month 06
day 30
svn rev 72865
language R
version.string R version 3.4.1 (2017-06-30)
nickname Single Candle
--
~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.
___ Patrick Connolly
{~._.~} Great minds discuss ideas
_( Y )_ Average minds discuss events
(:_~*~_:) Small minds discuss people
(_)-(_) ..... Eleanor Roosevelt
~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.
