基于字符串的J48算法绘制决策树进行预测

Question

I'm trying to plot J48 decision tree based on string values attributes and predict the target variable(categorical), I have seen many examples plotting decision tree based on numerical values but i haven't come across based on strings.

Here is sample data set, J48 decision tree works fine.

library(RWeka)
library(party)

MyData2 <- read.csv(file="iris.csv", header=TRUE, sep=",")
m3 <- J48(species~ ., data = MyData2)`enter code here`
if(require("party", quietly = TRUE)) plot(m3)


sepal_length    sepal_width petal_length    petal_width     species
5.1           3.5             1.4             0.2            setosa
4.9           3               1.4             0.2            setosa
7             3.2             4.7             1.4            versicolor
6.4           3.2             4.5             1.5            versicolor
6.3           3.3             6               2.5            virginica
5.8           2.7             5.1             1.9            virginica

If i rename the header sepal_length, sepal_width and to sepal_color and have values as "white", "black" with different combinations of colors to setosa, versicolor and virginca, how do i plot decision tree and predict the target species value.

Suppose if i have data set like below,

 sepal_color    sepal_color petal_color petal_color species
    white         black       white        black    setosa
    white         yellow      white        yellow   versicolor
    green         brown       green        brown    virginica

Answer 1

If the string variables represent levels of a categorical variable, then they should be turned into a factor() in R. Then, J48() can deal with these appropriately (just like other regression functions).

However, if the strings contain free text, then these are not supported directly. A feature preprocessing to some numeric or factor variable would be necessary before calling J48() .

As an example for classification based on categorical variables, let's turn the variables in the iris data into factors with three levels low , medium , high (cutting each variable into three equally-sized groups at the corresponding quantiles):

## load data and convert to factors via cut()
data("iris", package = "datasets")
for(i in 1:4) iris[[i]] <- cut(iris[[i]],
  quantile(iris[[i]], 0:3/3),
  labels = c("low", "medium", "high")
)
head(iris, 3)
##   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
## 1          low        high          low         low  setosa
## 2          low      medium          low         low  setosa
## 3          low      medium          low         low  setosa

## fit and plot J4.8 tree
j48 <- J48(Species ~ ., data = iris)
plot(j48)

Answer 2

Does the algorithm allow string regressors? I tried it and it threw an error. With strings you could try one-hot encoding eg "White"=1; "Black"=2 etc. eg

MyData2 <- iris
MyData2$Colour <- 2
MyData2[MyData2$Species == "setosa", ]$Colour <- 1
m3 <- J48(Species~ ., data = MyData2)
plot(m3)