This question should be answered using the
Weekly data set, which is part of the ISLR2
package. This data is similar in nature to the Smarket data
from this chapter’s lab, except that it contains 1, 089 weekly returns
for 21 years, from the beginning of 1990 to the end of
2010.
library(ISLR2)
attach(Weekly)
(a) Produce some numerical and graphical summaries of the
Weekly data. Do there appear to be any
patterns?
summary(Weekly)
## Year Lag1 Lag2 Lag3
## Min. :1990 Min. :-18.1950 Min. :-18.1950 Min. :-18.1950
## 1st Qu.:1995 1st Qu.: -1.1540 1st Qu.: -1.1540 1st Qu.: -1.1580
## Median :2000 Median : 0.2410 Median : 0.2410 Median : 0.2410
## Mean :2000 Mean : 0.1506 Mean : 0.1511 Mean : 0.1472
## 3rd Qu.:2005 3rd Qu.: 1.4050 3rd Qu.: 1.4090 3rd Qu.: 1.4090
## Max. :2010 Max. : 12.0260 Max. : 12.0260 Max. : 12.0260
## Lag4 Lag5 Volume Today
## Min. :-18.1950 Min. :-18.1950 Min. :0.08747 Min. :-18.1950
## 1st Qu.: -1.1580 1st Qu.: -1.1660 1st Qu.:0.33202 1st Qu.: -1.1540
## Median : 0.2380 Median : 0.2340 Median :1.00268 Median : 0.2410
## Mean : 0.1458 Mean : 0.1399 Mean :1.57462 Mean : 0.1499
## 3rd Qu.: 1.4090 3rd Qu.: 1.4050 3rd Qu.:2.05373 3rd Qu.: 1.4050
## Max. : 12.0260 Max. : 12.0260 Max. :9.32821 Max. : 12.0260
## Direction
## Down:484
## Up :605
##
##
##
##
cor(Weekly[,-9])
## Year Lag1 Lag2 Lag3 Lag4
## Year 1.00000000 -0.032289274 -0.03339001 -0.03000649 -0.031127923
## Lag1 -0.03228927 1.000000000 -0.07485305 0.05863568 -0.071273876
## Lag2 -0.03339001 -0.074853051 1.00000000 -0.07572091 0.058381535
## Lag3 -0.03000649 0.058635682 -0.07572091 1.00000000 -0.075395865
## Lag4 -0.03112792 -0.071273876 0.05838153 -0.07539587 1.000000000
## Lag5 -0.03051910 -0.008183096 -0.07249948 0.06065717 -0.075675027
## Volume 0.84194162 -0.064951313 -0.08551314 -0.06928771 -0.061074617
## Today -0.03245989 -0.075031842 0.05916672 -0.07124364 -0.007825873
## Lag5 Volume Today
## Year -0.030519101 0.84194162 -0.032459894
## Lag1 -0.008183096 -0.06495131 -0.075031842
## Lag2 -0.072499482 -0.08551314 0.059166717
## Lag3 0.060657175 -0.06928771 -0.071243639
## Lag4 -0.075675027 -0.06107462 -0.007825873
## Lag5 1.000000000 -0.05851741 0.011012698
## Volume -0.058517414 1.00000000 -0.033077783
## Today 0.011012698 -0.03307778 1.000000000
The only variables that appear to have any relation are Volume and Year.
(b) Use the full data set to perform a logistic regression
with Direction as the response and the five lag variables
plus Volume as predictors. Use the summary function to
print the results. Do any of the predictors appear to be statistically
significant? If so, which ones?
glm.fits<-glm(Direction~Lag1+Lag2+Lag3+Lag4+Lag5+Volume, data=Weekly,family=binomial)
summary(glm.fits)
##
## Call:
## glm(formula = Direction ~ Lag1 + Lag2 + Lag3 + Lag4 + Lag5 +
## Volume, family = binomial, data = Weekly)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.6949 -1.2565 0.9913 1.0849 1.4579
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 0.26686 0.08593 3.106 0.0019 **
## Lag1 -0.04127 0.02641 -1.563 0.1181
## Lag2 0.05844 0.02686 2.175 0.0296 *
## Lag3 -0.01606 0.02666 -0.602 0.5469
## Lag4 -0.02779 0.02646 -1.050 0.2937
## Lag5 -0.01447 0.02638 -0.549 0.5833
## Volume -0.02274 0.03690 -0.616 0.5377
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 1496.2 on 1088 degrees of freedom
## Residual deviance: 1486.4 on 1082 degrees of freedom
## AIC: 1500.4
##
## Number of Fisher Scoring iterations: 4
The only variable that appears to be statistically significant is Lag2.
(c) Compute the confusion matrix and overall fraction of correct predictions. Explain what the confusion matrix is telling you about the types of mistakes made by logistic regression.
glm.probs<-predict(glm.fits, type="response")
glm.pred<-rep("Down",1089)
glm.pred[glm.probs > 0.5] = "Up"
table(glm.pred, Direction)
## Direction
## glm.pred Down Up
## Down 54 48
## Up 430 557
\(\frac{54+557}{54+48+430+557}=0.5611\) shows
that 56.11% of the time the Direction was correctly predicted.
\(\frac{557}{48+557}=0.9207\) means Ups
were correctly predicted 92.07% of the time.
\(\frac{54}{54+430}=0.1116\) means
Downs were only correctly predicted 11.16% of the time.
(d) Now fit the logistic regression model using a training
data period from 1990 to 2008, with Lag2 as the only
predictor. Compute the confusion matrix and the overall fraction of
correct predictions for the held out data (that is, the data from 2009
and 2010).
train<-(Year<2009)
Weekly.910 <-Weekly[!train,]
Direction.910<-Direction[!train]
glm.fits<-glm(Direction~Lag2, data=Weekly, family=binomial, subset=train)
glm.probs<-predict(glm.fits, Weekly.910, type="response")
glm.pred<-rep("Down", length(glm.probs))
glm.pred[glm.probs > 0.5] = "Up"
table(glm.pred, Direction.910)
## Direction.910
## glm.pred Down Up
## Down 9 5
## Up 34 56
mean(glm.pred == Direction.910)
## [1] 0.625
By making a training subset with Lag2 as the predictor, we correctly
predicted the Direction 62.5% of the time. \(\frac{56}{5+56}=0.9180\) means Ups were
correctly predicted 91.8% of the time.
\(\frac{9}{9+34}=0.2093\) means Downs
were only correctly predicted 20.93% of the time. This is an improvement
from the previous model.
(e) Repeat (d) using LDA.
library(MASS)
##
## Attaching package: 'MASS'
## The following object is masked from 'package:ISLR2':
##
## Boston
lda.fits<-lda(Direction~Lag2, data=Weekly, family=binomial, subset=train)
lda.pred<-predict(lda.fits, Weekly.910)
table(lda.pred$class, Direction.910)
## Direction.910
## Down Up
## Down 9 5
## Up 34 56
mean(lda.pred$class==Direction.910)
## [1] 0.625
The LDA produces similar results to the logistic model in (d).
(f) Repeat (d) using QDA.
qda.fits<-qda(Direction~Lag2, data=Weekly, subset=train)
qda.pred<-predict(qda.fits, Weekly.910)$class
table(qda.pred, Direction.910)
## Direction.910
## qda.pred Down Up
## Down 0 0
## Up 43 61
mean(qda.pred==Direction.910)
## [1] 0.5865385
The QDA predicted everything as an Up, which has worse overall results than the previous models.
(g) Repeat (d) using KNN with K = 1.
library(class)
train.X<-as.matrix(Lag2[train])
test.X<-as.matrix(Lag2[!train])
train.Direction<-Direction[train]
set.seed(1)
knn.pred<-knn(train.X, test.X, train.Direction, k=1)
table(knn.pred, Direction.910)
## Direction.910
## knn.pred Down Up
## Down 21 30
## Up 22 31
mean(knn.pred == Direction.910)
## [1] 0.5
The KNN model appears to be entirely random, with a 50% success rate.
(h) Repeat (d) using naive Bayes.
library(e1071)
nb.fits<-naiveBayes(Direction~Lag2, data=Weekly, subset=train)
nb.pred<-predict(nb.fits, Weekly.910)
table(nb.pred, Direction.910)
## Direction.910
## nb.pred Down Up
## Down 0 0
## Up 43 61
mean(nb.pred == Direction.910)
## [1] 0.5865385
Naive Bayes also predicts everything as Up.
(i) Which of these methods appears to provide the best results on this data?
The logistic regression and the LDA appears to have the best results on the data.
(j) Experiment with different combinations of predictors, including possible transformations and interactions, for each of the methods. Report the variables, method, and associated confusion matrix that appears to provide the best results on the held out data. Note that you should also experiment with values for K in the KNN classifier.
glm.fits<-glm(Direction~Lag2+Lag4, data=Weekly, family=binomial, subset=train)
glm.probs<-predict(glm.fits, Weekly.910, type="response")
glm.pred<-rep("Down", length(glm.probs))
glm.pred[glm.probs > 0.5] = "Up"
table(glm.pred, Direction.910)
## Direction.910
## glm.pred Down Up
## Down 8 4
## Up 35 57
mean(glm.pred == Direction.910)
## [1] 0.625
lda.fits<-lda(Direction~Lag2+Lag3+Lag4, data=Weekly, family=binomial, subset=train)
lda.pred<-predict(lda.fits, Weekly.910)
table(lda.pred$class, Direction.910)
## Direction.910
## Down Up
## Down 8 5
## Up 35 56
mean(lda.pred$class==Direction.910)
## [1] 0.6153846
qda.fits<-qda(Direction~Lag2+Lag5, data=Weekly, subset=train)
qda.pred<-predict(qda.fits, Weekly.910)$class
table(qda.pred, Direction.910)
## Direction.910
## qda.pred Down Up
## Down 3 11
## Up 40 50
mean(qda.pred==Direction.910)
## [1] 0.5096154
knn.pred<-knn(train.X, test.X, train.Direction, k=3)
table(knn.pred, Direction.910)
## Direction.910
## knn.pred Down Up
## Down 16 19
## Up 27 42
mean(knn.pred == Direction.910)
## [1] 0.5576923
nb.fits<-naiveBayes(Direction~Lag1+Lag2, data=Weekly, subset=train)
nb.pred<-predict(nb.fits, Weekly.910)
table(nb.pred, Direction.910)
## Direction.910
## nb.pred Down Up
## Down 3 8
## Up 40 53
mean(nb.pred == Direction.910)
## [1] 0.5384615
detach(Weekly)
Despite trying many different transformations using multiple predictors, nothing was able to surpass the original logistic regression and LDA that used just Lag2.
In this problem, you will develop a model to predict whether a given car gets high or low gas mileage based on the Auto data set.
attach(Auto)
(a) Create a binary variable, mpg01 , that
contains a 1 if mpg contains a value above its median, and
a 0 if mpg contains a value below its median. You can
compute the median using the median() function. Note you
may find it helpful to use the data.frame() function to
create a single data set containing both mpg01 and the
other Auto variables.
mpg01<-rep(0, length(mpg))
mpg01[mpg > median(mpg)]<-1
Auto01<-data.frame(Auto, mpg01)
rm(mpg01)
(b) Explore the data graphically in order to investigate the
association between mpg01 and the other features. Which of
the other features seem most likely to be useful in predicting
mpg01? Scatterplots and boxplots may be useful tools to
answer this question. Describe your findings.
pairs(Auto01[,1:10])
Displacement, horsepower, and weight appear to correlate with mpg01.
(c) Split the data into a training set and a test set.
detach(Auto)
attach(Auto01)
train<-(year%%2==0)
train.X<-Auto01[train,]
test.X<-Auto01[-train,]
(d) Perform LDA on the training data in order to predict
mpg01 using the variables that seemed most associated with
mpg01 in (b). What is the test error of the model
obtained?
lda.fits<-lda(mpg01~displacement+horsepower+weight, data=train.X)
lda.pred<-predict(lda.fits, test.X)
table(lda.pred$class, test.X$mpg01)
##
## 0 1
## 0 160 11
## 1 35 185
mean(lda.pred$class != test.X$mpg01)
## [1] 0.1176471
The LDA has a test error rate of 11.76%.
(e) Perform QDA on the training data in order to predict
mpg01 using the variables that seemed most associated with
mpg01 in (b). What is the test error of the model
obtained?
qda.fits<-qda(mpg01~displacement+horsepower+weight, data=train.X)
qda.pred<-predict(qda.fits, test.X)
table(qda.pred$class, test.X$mpg01)
##
## 0 1
## 0 170 17
## 1 25 179
mean(qda.pred$class != test.X$mpg01)
## [1] 0.1074169
The QDA has a test error rate of 10.74%.
(f) Perform logistic regression on the training data in order
to predict mpg01 using the variables that seemed most
associated with mpg01 in (b). What is the test error of the
model obtained?
glm.fits<-glm(mpg01~displacement+horsepower+weight, data=train.X, family=binomial)
glm.probs<-predict(glm.fits, test.X, type="response")
glm.pred<-rep(0, length(glm.probs))
glm.pred[glm.probs > 0.5] = 1
table(glm.pred, test.X$mpg01)
##
## glm.pred 0 1
## 0 172 16
## 1 23 180
mean(glm.pred != test.X$mpg01)
## [1] 0.09974425
The logistic regression has a test error rate of 9.97%.
(g) Perform naive Bayes on the training data in order to
predict mpg01 using the variables that seemed most
associated with mpg01 in (b). What is the test error of the
model obtained?
nb.fits<-naiveBayes(mpg01~displacement+horsepower+weight, data=train.X)
nb.pred<-predict(nb.fits, test.X)
table(nb.pred, test.X$mpg01)
##
## nb.pred 0 1
## 0 169 14
## 1 26 182
mean(nb.pred != test.X$mpg01)
## [1] 0.1023018
The Naive Bayes has a test error rate of 10.23%.
(h) Perform KNN on the training data, with several values of
K, in order to predict mpg01 . Use only the variables that
seemed most associated with mpg01 in (b). What test errors
do you obtain? Which value of K seems to perform the best on this data
set?
train.K<-cbind(displacement, horsepower, weight)[train,]
test.K<-cbind(displacement, horsepower, weight)[-train,]
knn.pred<-knn(train.K,test.K,train.X$mpg01,k=1)
mean(knn.pred != test.X$mpg01)
## [1] 0.07161125
knn.pred<-knn(train.K,test.K,train.X$mpg01,k=3)
mean(knn.pred != test.X$mpg01)
## [1] 0.0971867
knn.pred<-knn(train.K,test.K,train.X$mpg01,k=10)
mean(knn.pred != test.X$mpg01)
## [1] 0.1202046
detach(Auto01)
k=1 appears to be the best value of K on this data set, with a 7.16% test error rate.
Using the Boston data set, fit classification
models in order to predict whether a given census tract has a crime rate
above or below the median. Explore logistic regression, LDA, naive
Bayes, and KNN models using various subsets of the predictors. Describe
your findings.
Hint: You will have to create the response variable
yourself, using the variables that are contained in the
Boston data set.
attach(Boston)
crim01<-rep(0, length(crim))
crim01[crim > median(crim)]<-1
Boston01<-data.frame(Boston, crim01)
rm(crim01)
detach(Boston)
attach(Boston01)
train<-1:(dim(Boston01)[1]/2)
test<-(dim(Boston01)[1]/2 + 1):dim(Boston01)[1]
train.Boston = Boston01[train,]
test.Boston = Boston01[test,]
test.crim = crim01[test]
glm.fits<-glm(crim01~indus+nox+age+dis+rad+tax, data=train.Boston, family=binomial)
glm.probs = predict(glm.fits, test.Boston, type = "response")
glm.pred = rep(0, length(glm.probs))
glm.pred[glm.probs > 0.5] = 1
table(glm.pred, test.crim)
## test.crim
## glm.pred 0 1
## 0 75 8
## 1 15 155
mean(glm.pred != test.crim)
## [1] 0.09090909
lda.fits<-lda(crim01~indus+nox+age+dis+rad+tax, data=train.Boston)
lda.pred<-predict(lda.fits, test.Boston)
table(lda.pred$class, test.Boston$crim01)
##
## 0 1
## 0 81 18
## 1 9 145
mean(lda.pred$class != test.Boston$crim01)
## [1] 0.1067194
nb.fits<-naiveBayes(crim01~indus+nox+age+dis+rad+tax, data=train.Boston)
nb.pred<-predict(nb.fits, test.Boston)
table(nb.pred, test.Boston$crim01)
##
## nb.pred 0 1
## 0 77 13
## 1 13 150
mean(nb.pred != test.Boston$crim01)
## [1] 0.1027668
set.seed(1)
train.K<-cbind(indus, nox, age, dis, rad, tax)[train,]
test.K<-cbind(indus, nox, age, dis, rad, tax)[-train,]
knn.pred<-knn(train.K,test.K,train.Boston$crim01,k=1)
table(knn.pred, test.Boston$crim01)
##
## knn.pred 0 1
## 0 82 151
## 1 8 12
mean(knn.pred != test.Boston$crim01)
## [1] 0.6284585
knn.pred<-knn(train.K,test.K,train.Boston$crim01,k=5)
table(knn.pred, test.Boston$crim01)
##
## knn.pred 0 1
## 0 80 22
## 1 10 141
mean(knn.pred != test.Boston$crim01)
## [1] 0.1264822
knn.pred<-knn(train.K,test.K,train.Boston$crim01,k=10)
table(knn.pred, test.Boston$crim01)
##
## knn.pred 0 1
## 0 83 23
## 1 7 140
mean(knn.pred != test.Boston$crim01)
## [1] 0.1185771
knn.pred<-knn(train.K,test.K,train.Boston$crim01,k=15)
table(knn.pred, test.Boston$crim01)
##
## knn.pred 0 1
## 0 83 23
## 1 7 140
mean(knn.pred != test.Boston$crim01)
## [1] 0.1185771
knn.pred<-knn(train.K,test.K,train.Boston$crim01,k=20)
table(knn.pred, test.Boston$crim01)
##
## knn.pred 0 1
## 0 83 23
## 1 7 140
mean(knn.pred != test.Boston$crim01)
## [1] 0.1185771
The generalized logistic model is the best performing of the models in this scenario, with a 9.09% test error rate. Other models tested were mostly just over 10%, though the k=1 KNN model had a massive 62.85% error rate. The indus, nox, age, dis, rad, and tax variables were all predictors of whether crime rates were above or below the median.