hw4q2.md

In MickyDowns/mine_algorithms: What the package does (short line)

hw4q2 - KNN on sonar data

q2a

Load data, libs

setwd("./data")

train<-read.csv("sonar_train.csv",header=F)
test<-read.csv("sonar_test.csv",header=F)

setwd("../")

library(class)

Create training and test sets

# create x and y training vars
y<-as.factor(train[,61])
x<-train[,1:60]

# create x and y test vars
y_test<-as.factor(test[,61])
x_test<-test[,1:60]

Fit models varying k. Assess fit on train and test data.

iter<-10

errors<-matrix(data=NA,nrow=iter,ncol=3)
colnames(errors)<-c("k","train","test")

for(i in 1:iter){
     errors[i,1]<-i

     # fit model on training data
     fit.tr<-knn(x,x,cl=y,k=i)

     # calculate/log training error
     errors[i,2]<-1-sum(y==fit.tr)/length(y)

     # fit model on test data
     fit.te<-knn(x,x_test,y,k=i)

     # calculate/log test error
     errors[i,3]<-1-sum(y_test==fit.te)/length(y_test)
}

Output graph

plot(errors[,2]~errors[,1],type="l",col="blue",
     main="Sonar KNN Train & Test Errors (Michael Downs)",
     xlab="k",ylab="error rate",ylim=c(0,0.4),lwd=3)

lines(errors[,3]~errors[,1],type="l",col="red4",lwd=3)

legend("bottomright",col=c("blue","red4"),lwd=3,
       legend=c("training error","test error"))

Answer question

minVal<-errors[which.min(errors[,3]),3]
minK<-errors[which.min(errors[,3]),1]
print(c("Lowest test error of ",round(minVal,3),"and k of ",minK))

##                                            test                         
## "Lowest test error of "                 "0.167"             "and k of " 
##                       k 
##                     "2"

q2b

Save off errors

errors1<-errors

Load data, libs

setwd("./data")

train<-read.csv("sonar_train.csv",header=F)
test<-read.csv("sonar_test.csv",header=F)

setwd("../")

library(class)

Create training and test sets

# create x and y training vars
y<-as.factor(train[,61])
x<-train[,1:60]

# create x and y test vars
y_test<-as.factor(test[,61])
x_test<-test[,1:60]

Fit models varying k. Assess fit on train and test data.

iter<-10

errors<-matrix(data=NA,nrow=iter,ncol=3)
colnames(errors)<-c("k","train","test")

for(i in 1:iter){
     errors[i,1]<-i

     # fit model on training data
     fit.tr<-knn(x,x,cl=y,k=i)

     # calculate/log training error
     errors[i,2]<-1-sum(y==fit.tr)/length(y)

     # fit model on test data
     fit.te<-knn(x,x_test,y,k=i)

     # calculate/log test error
     errors[i,3]<-1-sum(y_test==fit.te)/length(y_test)
}

Plot both sets of errors

plot(errors[,2]~errors[,1],type="l",col="blue",
     main="Sonar KNN Train & Test Errors (Michael Downs)",
     xlab="k",ylab="error rate",ylim=c(0,0.4),lwd=3)

lines(errors[,3]~errors[,1],type="l",col="red1",lwd=3)

lines(errors1[,2]~errors1[,1],type="l",col="light blue",lwd=3,lty=2)
lines(errors1[,3]~errors1[,1],type="l",col="red4",lwd=3,lty=2)

legend("bottomright",col=c("blue","lightblue","red4","red1"),lwd=3,
       legend=c("train error iter2","train error iter1", "test error iter2","test error iter1"))

Answer question

minVal<-errors[which.min(errors[,3]),3]
minK<-errors[which.min(errors[,3]),1]
print(c("Lowest test error of ",round(minVal,3),"and k of ",minK))

##                                            test                         
## "Lowest test error of "                 "0.167"             "and k of " 
##                       k 
##                     "4"

MickyDowns/mine_algorithms documentation built on May 8, 2019, 10:49 a.m.