Tree

#install.packages("ISLR") # if you don't have the package install it.
#install.packages("tree")# if you don't have the package install it.
library(ISLR)
library(tree)
attach(Carseats)
range(Sales)
high = ifelse(Sales >=8, "YES", "NO")
Carseats=data.frame(Carseats, high)
range(Sales)
set.seed(2)
str(Carseats)
range(high)
names(Carseats)
Carseats <- Carseats[,-1]
train <- sample(1:nrow(Carseats), nrow(Carseats)/2)
test = -train
training_data= Carseats[train,]
testing_data = Carseats[test,]
testing_high <- high[test]
tree_model <- tree(high~., training_data)
plot(tree_model)
text(tree_model,pretty=0)
#Predict
tree_pred = predict(tree_model, testing_data, type="class")
#Check model accuracy
error_rate = mean(tree_pred != testing_high) # 28%

##Pruning the tree. Cross-validate to see where to stop pruning
set.seed(9)
cv_tree = cv.tree(tree_model, FUN=prune.misclass)
cv_tree
names(cv_tree) #dev = cross_validation error rate  
plot(cv_tree$size,cv_tree$dev, type = "b")
#Pruned model
pruned_model = prune.misclass(tree_model, best=9)
plot(pruned_model)
text(pruned_model)
#Check accuracy
tree_pred2=predict(pruned_model, testing_data, type = "class")
error_rate2= mean(tree_pred2 != testing_high) # 23%

R 3.2.2 Error

I got this peculiar error. After installing ggplot2, i couldn't use it. R said there is no such package.
Solution: install package:‘Rcpp’
install.packages("Rcpp")

Subsetting

# select 1 through 9th row, all columns
iris[1:9,]
# select 1st column through 4th column, all rows
iris[,1:4]
# calculate column sums
colSums(iris[,1:4])
rowSums(iris[1:2,1:4])
dim(iris) #150R x 5C

iris[c('Sepal.Length','Petal.Length')]
#iris$Sepal.Length<5 creates a vector of TRUE & FALSES
iris[iris$Sepal.Length<5,]

#redo  -- Same as previous code snippet
tf <- iris$Sepal.Length<5
iris[tf,]
str(th)
#redo with subset
subset(iris,Sepal.Length<5)

#random selection of sample
x <- sample(1:nrow(iris), 0.33 * nrow(iris),replace=F)
iris[x,]

GGPlot2 Lesson1

Ref: http://www.statmethods.net/advgraphs/ggplot2.html
library(ggplot2)
 qplot(iris$Sepal.Length,iris$Sepal.Width)


 qplot(iris$Sepal.Length,iris$Sepal.Width, col=iris$Species)

R fellowship

https://projects.propublica.org/data-store/

Cluster Analysis

#K-means
par(mfrow=c(1,2))
x=iris$Sepal.Length
y=iris$Sepal.Width
s=iris$Species
kdf <- data.frame(x,y)
km <- kmeans(kdf[,c(1,2)],5)
plot(x,y, col=km$cluster, cex=1,pch=1, main="Plot of iris k-means Clusters")
points(km$centers, cex=1, pch=3, lwd=3)
plot(x,y, col=s, cex=1,pch=1, main="Plot of iris")


#Hierarchical Clustering
x=iris$Sepal.Length
y=iris$Sepal.Width
kdf <- data.frame(x,y)
disto=dist(kdf)
hc=hclust(disto)
plot(hc)

Resource: http://www.rdatamining.com/examples/kmeans-clustering

Frequency Table

Here is the data from Google search of continents (all figures in millions):
North America,577
South America,453
Europe,1840
Asia,1510
Africa,1620

We use the concatenation operator to create the data set:

> continents <- c(rep("North America",577),
+                 rep("South America",453),
+                 rep("Europe",1840),
+                 rep("Asia",1510),
+                 rep("Africa",1620))

Next, we use the table function to create the frequencies.

> table(continents)

Following is the output:

But, the frequencies are not ordered according to the magnitude. We can easily do that with sort function: