Skip to content

Commit 2ffd046

Browse files
ariuttaariutta
committed
Added ignore file for git.
1 parent 65ace18 commit 2ffd046

4 files changed

Lines changed: 111 additions & 67 deletions

File tree

‎.gitignore‎

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,3 @@
1+
.Rhistory
2+
Icon*
3+
.RData

‎8_PREDMACHLEARN/Practical Machine Learning Course Notes.Rmd‎

Lines changed: 42 additions & 20 deletions
Original file line numberDiff line numberDiff line change
@@ -2,13 +2,13 @@
22
title: "Practical Machine Learning Course Notes"
33
author: "Xing Su"
44
output:
5-
pdf_document:
6-
toc: yes
7-
toc_depth: 3
85
html_document:
96
highlight: pygments
107
theme: spacelab
118
toc: yes
9+
pdf_document:
10+
toc: yes
11+
toc_depth: 3
1212
header-includes: \usepackage{graphicx} \usepackage{mathtools}
1313
---
1414

@@ -22,8 +22,11 @@ $\pagebreak$
2222
- ***Note**: choosing the right dataset and knowing what the specific question is are paramount to the success of the prediction algorithm (GoogleFlu failed to predict accurately when people's search habits changed) *
2323

2424
```{r fig.width = 4, fig.height = 3, fig.align = 'center', echo = FALSE, message = FALSE}
25-
library(png);library(grid)
26-
library(doMC); registerDoMC(cores = 4)
25+
# install grid and png packages if not present
26+
if (!require("pacman")) install.packages("pacman")
27+
p_load("png", "grid")
28+
p_load("doMC")
29+
registerDoMC(cores = 4)
2730
grid.raster(readPNG("figures/1.png"))
2831
```
2932

@@ -65,7 +68,8 @@ grid.raster(readPNG("figures/1.png"))
6568

6669
```{r message = FALSE, fig.width = 3, fig.height = 3, fig.align = 'center'}
6770
# load data
68-
library(kernlab); data(spam); set.seed(333)
71+
p_load("kernlab")
72+
data(spam); set.seed(333)
6973
# picking a small subset (10 values) from spam data set
7074
smallSpam <- spam[sample(dim(spam)[1],size=10),]
7175
# label spam = 2 and ham = 1
@@ -352,7 +356,7 @@ grid.raster(readPNG("figures/13.png"))
352356

353357
```{r message = FALSE}
354358
# load packages and data
355-
library(caret)
359+
p_load("caret")
356360
# create training set indexes with 75% of data
357361
inTrain <- createDataPartition(y=spam$type,p=0.75, list=FALSE)
358362
# subset spam data to training
@@ -487,7 +491,7 @@ args(trainControl)
487491

488492
```{r fig.width = 5, fig.height = 5, fig.align = 'center'}
489493
# load relevant libraries
490-
library(ISLR); library(ggplot2);
494+
p_load("ISLR", "ggplot2")
491495
# load wage data
492496
data(Wage)
493497
# create training and test sets
@@ -508,15 +512,15 @@ qplot(age,wage,colour=education,data=training)+geom_smooth(method='lm',formula=y
508512
```
509513

510514
* `cut2(variable, g=3)` = creates a new factor variable by cutting the specified variable into n groups (3 in this case) based on percentiles
511-
- ***Note**: `cut2` function is part of the `Hmisc` package, so `library(Hmisc)` must be run first *
515+
- ***Note**: `cut2` function is part of the `Hmisc` package, so `p_load("Hmisc")` must be run first *
512516
- this variable can then be used to tabulate/plot the data
513517
* `grid.arrange(p1, p2, ncol=2)` = `ggplot2` function the print multiple graphs on the same plot
514-
- ***Note**: `grid.arrange` function is part of the `gridExtra` package, so `library(gridExtra)` must be run first *
518+
- ***Note**: `grid.arrange` function is part of the `gridExtra` package, so `p_load("gridExtra")` must be run first *
515519

516520
```{r fig.width = 7, fig.height = 3, fig.align = 'center', message = FALSE}
517521
# load Hmisc and gridExtra packages
518-
library(Hmisc);library(gridExtra);
519-
# cute the wage variable
522+
p_load("Hmisc", "gridExtra")
523+
# cut the wage variable
520524
cutWage <- cut2(training$wage,g=3)
521525
# plot the boxplot
522526
p1 <- qplot(cutWage,age, data=training,fill=cutWage,
@@ -595,6 +599,7 @@ rbind(train = c(mean = mean(trainCapAveS), std = sd(trainCapAveS)),
595599

596600
```{r message = FALSE, fig.width = 5, fig.height = 3, fig.align = 'center'}
597601
# set up BoxCox transforms
602+
p_load("e1071")
598603
preObj <- preProcess(training[,-58],method=c("BoxCox"))
599604
# perform preprocessing on training data
600605
trainCapAveS <- predict(preObj,training[,-58])$capitalAve
@@ -610,6 +615,7 @@ par(mfrow=c(1,2)); hist(trainCapAveS); qqnorm(trainCapAveS)
610615

611616
```{r message = FALSE}
612617
# Make some values NA
618+
p_load("RANN")
613619
training$capAve <- training$capitalAve
614620
selectNA <- rbinom(dim(training)[1],size=1,prob=0.05)==1
615621
training$capAve[selectNA] <- NA
@@ -691,7 +697,7 @@ nearZeroVar(training,saveMetrics=TRUE)
691697

692698
```{r fig.width = 4, fig.height = 3, fig.align = 'center'}
693699
# load splines package
694-
library(splines)
700+
p_load("splines")
695701
# create polynomial function
696702
bsBasis <- bs(training$age,df=3)
697703
# fit the outcome on the three polynomial terms
@@ -767,6 +773,7 @@ plot(prComp$x[,1],prComp$x[,2],col=typeColor,xlab="PC1",ylab="PC2")
767773

768774
```{r message = FALSE, warning = FALSE, fig.width = 4, fig.height = 3, fig.align = 'center'}
769775
# create train and test sets
776+
p_load("caret")
770777
inTrain <- createDataPartition(y=spam$type,p=0.75, list=FALSE)
771778
training <- spam[inTrain,]
772779
testing <- spam[-inTrain,]
@@ -775,7 +782,8 @@ preProc <- preProcess(log10(training[,-58]+1),method="pca",pcaComp=2)
775782
# calculate PCs for training data
776783
trainPC <- predict(preProc,log10(training[,-58]+1))
777784
# run model on outcome and principle components
778-
modelFit <- train(training$type ~ .,method="glm",data=trainPC)
785+
modelFit <- train(training$type ~.,data=trainPC,method="glm")
786+
779787
# calculate PCs for test data
780788
testPC <- predict(preProc,log10(testing[,-58]+1))
781789
# compare results
@@ -1003,6 +1011,15 @@ modFit <- train(Species ~ .,method="rpart",data=training)
10031011
# print the classification tree
10041012
print(modFit$finalModel)
10051013
# plot the classification tree
1014+
# Note: Need to install Gtk+ 2.x before using rattle. On OS/X:
1015+
# brew install gtk+
1016+
# I should just need to run the following:
1017+
#p_load("rattle")
1018+
# But I actually need to run this, because apparently the dependencies aren't loading:
1019+
p_load("RGtk2")
1020+
p_load("rpart")
1021+
p_load("rpart.plot")
1022+
p_load("rattle")
10061023
rattle::fancyRpartPlot(modFit$finalModel)
10071024
# predict on test values
10081025
predict(modFit,newdata=testing)
@@ -1024,7 +1041,8 @@ predict(modFit,newdata=testing)
10241041

10251042
```{r fig.width = 5, fig.height = 4, fig.align = 'center', message = FALSE, warning = FALSE, cache = FALSE}
10261043
# load data
1027-
library(ElemStatLearn); data(ozone,package="ElemStatLearn")
1044+
p_load("ElemStatLearn")
1045+
data(ozone,package="ElemStatLearn")
10281046
# reorder rows based on ozone variable
10291047
ozone <- ozone[order(ozone$ozone),]
10301048
# create empty matrix
@@ -1065,7 +1083,8 @@ lines(1:155,apply(ll,2,mean),col="red",lwd=2)
10651083

10661084
```{r fig.width = 5, fig.height = 4, fig.align = 'center', message = FALSE, warning = FALSE, cache = FALSE}
10671085
# load relevant package and data
1068-
library(party); data(ozone,package="ElemStatLearn")
1086+
p_load("party")
1087+
data(ozone,package="ElemStatLearn")
10691088
# reorder rows based on ozone variable
10701089
ozone <- ozone[order(ozone$ozone),]
10711090
# extract predictors
@@ -1128,6 +1147,7 @@ grid.raster(readPNG("figures/14.png"))
11281147
* ***example***
11291148

11301149
```{r fig.width = 5, fig.height = 3, fig.align = 'center', message = FALSE, warning = FALSE, cache = FALSE}
1150+
p_load("randomForest")
11311151
# load data
11321152
data(iris)
11331153
# create train/test data sets
@@ -1208,6 +1228,7 @@ grid.raster(readPNG("figures/18.png"))
12081228
* ***example***
12091229

12101230
```{r message = FALSE, warning = FALSE, cache = FALSE}
1231+
p_load("gbm")
12111232
# load data
12121233
data(Wage)
12131234
# remove log wage variable (we are trying to predict wage)
@@ -1321,6 +1342,7 @@ pred.lda
13211342
- ***example: `caret` package***
13221343

13231344
```{r message = F, warning = F}
1345+
p_load("klaR")
13241346
# using the same data from iris, run naive Bayes on training data
13251347
nb <- train(Species ~ ., data=training,method="nb")
13261348
# predict test outcomes using naive Bayes model
@@ -1541,7 +1563,7 @@ lm(lpsa ~ .,data =small)
15411563

15421564
```{r echo = FALSE, fig.width = 5, fig.height = 5, fig.align = 'center', message = FALSE, warning = FALSE}
15431565
# load MASS library for lm.ridge function
1544-
library(MASS)
1566+
p_load("MASS")
15451567
# we will run the models for 10 different values from 0 to 50
15461568
lambdas <- seq(0,50,len=10)
15471569
# create empty vectors for training and test residual sum squares
@@ -1640,7 +1662,7 @@ legend("topright",covnames,pch=as.character(1:8), cex = 0.5)
16401662

16411663
```{r fig.width = 5, fig.height = 5, fig.align = 'center', message = F, warning = F, results = 'hide'}
16421664
# load lars package
1643-
library(lars)
1665+
p_load("lars")
16441666
# perform lasso regression
16451667
lasso.fit <- lars(as.matrix(x), y, type="lasso", trace=TRUE)
16461668
# plot lasso regression model
@@ -1820,7 +1842,7 @@ $\pagebreak$
18201842

18211843
```{r fig.width = 5, fig.height = 4.5, fig.align = 'center', message = FALSE, warning = FALSE, cache = FALSE}
18221844
# load quantmod package
1823-
library(quantmod);
1845+
p_load("quantmod");
18241846
# specify to and from dates
18251847
from.dat <- as.Date("01/01/00", format="%m/%d/%y")
18261848
to.dat <- as.Date("3/2/15", format="%m/%d/%y")
@@ -1838,7 +1860,7 @@ plot(decompose(ts),xlab="Years")
18381860

18391861
```{r fig.width = 5, fig.height = 3.5, fig.align = 'center', message = FALSE, warning = FALSE}
18401862
# load forecast library
1841-
library(forecast)
1863+
p_load("forecast")
18421864
# find the number of rows (years)
18431865
rows <- ceiling(length(ts)/12)
18441866
# use 90% of the data to create training set

0 commit comments

Comments
 (0)