Tree-Based Methods for Predicting Authorship
#install.packages('tree')
library (tree)
Loading Training and Testing Data
library (ISLR)
attach (Carseats )
#head(Carseats)
df<-read.csv("author_training.csv")
kf<-read.csv("author_testing.csv")
Cnverting Response veriable to Numeric Class
library(dplyr)
##
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
##
## filter, lag
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union
mapping <- c("Austen" = 1, "London" = 2, "Milton" = 3, "Shakespeare" = 4)
df$Author.r <- mapping[df$Author]
kf$Author.r<- mapping[kf$Author]
Fitting Classification Tree Model
library(tree)
tree1 =tree(Author.r~.-Author,data=df)
summary(tree1)
##
## Regression tree:
## tree(formula = Author.r ~ . - Author, data = df)
## Variables actually used in tree construction:
## [1] "was" "my" "its" "be" "to" "this"
## Number of terminal nodes: 7
## Residual mean deviance: 0.06852 = 39.88 / 582
## Distribution of residuals:
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## -1.71400 -0.04072 0.01026 0.00000 0.01026 2.01000
tree1
## node), split, n, deviance, yval
## * denotes terminal node
##
## 1) root 589 725.800 2.085
## 2) was < 10.5 151 50.370 3.689
## 4) my < 9.5 39 16.970 2.974
## 8) its < 0.5 32 4.875 3.188 *
## 9) its > 0.5 7 4.000 2.000 *
## 5) my > 9.5 112 6.562 3.938 *
## 3) was > 10.5 438 153.100 1.532
## 6) be < 9.5 195 9.979 1.990 *
## 7) be > 9.5 243 69.420 1.165
## 14) to < 36.5 22 23.320 2.409
## 28) this < 9.5 15 2.400 1.800 *
## 29) this > 9.5 7 3.429 3.714 *
## 15) to > 36.5 221 8.633 1.041 *
plot(tree1)
text(tree1, pretty=0)
set.seed (2)
tree2 =tree(Author.r~.-Author ,data= df )
summary(tree2)
##
## Regression tree:
## tree(formula = Author.r ~ . - Author, data = df)
## Variables actually used in tree construction:
## [1] "was" "my" "its" "be" "to" "this"
## Number of terminal nodes: 7
## Residual mean deviance: 0.06852 = 39.88 / 582
## Distribution of residuals:
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## -1.71400 -0.04072 0.01026 0.00000 0.01026 2.01000
tree.pred=predict(tree2 ,kf)
#tree.pred
table(tree.pred,kf$Author.r)
##
## tree.pred 1 2 3 4
## 1.04072398190045 75 4 1 0
## 1.8 2 2 0 0
## 1.98974358974359 6 75 0 1
## 2 1 1 0 0
## 3.1875 1 0 12 5
## 3.71428571428571 0 1 0 0
## 3.9375 10 0 6 49
Error for Clssification Tree:
Error1= (75+78+12+49)/252
Error1
## [1] 0.8492063