EPPS 6323 Assignment 8

Linear Discriminant Analysis

# Knowledge Mining: Linear Discriminant Analysis
# File: Lab_LDA01.R
# Theme: Linear Discriminant Analysis
# Adapted from ISLR Chapter 4 Lab

# Setup
require(ISLR)
require(MASS)
require(descr)
attach(Smarket)

## Linear Discriminant Analysis
freq(Direction)

Direction 
      Frequency Percent
Down        602   48.16
Up          648   51.84
Total      1250  100.00

train = Year<2005
lda.fit=lda(Direction~Lag1+Lag2,data=Smarket, subset=Year<2005)
lda.fit

Call:
lda(Direction ~ Lag1 + Lag2, data = Smarket, subset = Year < 
    2005)

Prior probabilities of groups:
    Down       Up 
0.491984 0.508016 

Group means:
            Lag1        Lag2
Down  0.04279022  0.03389409
Up   -0.03954635 -0.03132544

Coefficients of linear discriminants:
            LD1
Lag1 -0.6420190
Lag2 -0.5135293

plot(lda.fit, col="dodgerblue")

Smarket.2005=subset(Smarket,Year==2005) # Creating subset with 2005 data for prediction
lda.pred=predict(lda.fit,Smarket.2005)
names(lda.pred)

[1] "class"     "posterior" "x"        

lda.class=lda.pred$class
Direction.2005=Smarket$Direction[!train] 
table(lda.class,Direction.2005) 

         Direction.2005
lda.class Down  Up
     Down   35  35
     Up     76 106

data.frame(lda.pred)[1:5,]

     class posterior.Down posterior.Up         LD1
999     Up      0.4901792    0.5098208  0.08293096
1000    Up      0.4792185    0.5207815  0.59114102
1001    Up      0.4668185    0.5331815  1.16723063
1002    Up      0.4740011    0.5259989  0.83335022
1003    Up      0.4927877    0.5072123 -0.03792892

table(lda.pred$class,Smarket.2005$Direction)

      
       Down  Up
  Down   35  35
  Up     76 106

mean(lda.pred$class==Smarket.2005$Direction)

[1] 0.5595238