Practical Machine Learning Final Project
Puxin Xu
January 22, 2016
library(lattice)
library(ggplot2)
library(caret)
library(rpart)
library(rpart.plot)
library(RColorBrewer)
library(rattle)
library(randomForest)
library(knitr)Background
Using devices such as Jawbone Up, Nike FuelBand, and Fitbit it is now possible to collect a large amount of data about personal activity relatively inexpensively. These type of devices are part of the quantified self movement – a group of enthusiasts who take measurements about themselves regularly to improve their health, to find patterns in their behavior, or because they are tech geeks. One thing that people regularly do is quantify how much of a particular activity they do, but they rarely quantify how well they do it. In this project, your goal will be to use data from accelerometers on the belt, forearm, arm, and dumbell of 6 participants. They were asked to perform barbell lifts correctly and incorrectly in 5 different ways. More information is available from the website here: http://groupware.les.inf.puc-rio.br/har (see the section on the Weight Lifting Exercise Dataset).
Data
The training data for this project are available here: https://d396qusza40orc.cloudfront.net/predmachlearn/pml-training.csv
The test data are available here: https://d396qusza40orc.cloudfront.net/predmachlearn/pml-testing.csv
The data for this project come from this source: http://groupware.les.inf.puc-rio.br/har . If you use the document you create for this class for any purpose please cite them as they have been very generous in allowing their data to be used for this kind of assignment.
Getting and Cleaning the data
Getting the data
if (!dir.exists("Final_Project")) {
dir.create("./Final_Project")}
setwd("./Final_Project")
fileUrl_training <- "https://d396qusza40orc.cloudfront.net/predmachlearn/pml-training.csv"
download.file(fileUrl_training, destfile = "./pml-training.csv")
fileUrl_testing <- "https://d396qusza40orc.cloudfront.net/predmachlearn/pml-testing.csv"
download.file(fileUrl_testing, destfile = "./pml-testing.csv")
training <- read.csv("pml-training.csv", na.strings=c("NA","#DIV/0!",""))
testing <- read.csv("pml-testing.csv", na.strings=c("NA","#DIV/0!",""))Cleaning the data
nzv <- nearZeroVar(training, saveMetrics=TRUE)
training <- training[,nzv$nzv==FALSE]
training <- training[c(-1)]
trainingtemp <- training
for(i in 1:length(training)) {
if( sum( is.na( training[, i] ) ) /nrow(training) >= .7) {
for(j in 1:length(trainingtemp)) {
if( length( grep(names(training[i]), names(trainingtemp)[j]) ) == 1) {
trainingtemp <- trainingtemp[ , -j]
}
}
}
}
training <- trainingtemp
rm(trainingtemp)
inTrain <- createDataPartition(training$classe, p=0.6, list=FALSE)
myTraining <- training[inTrain, ]
myTesting <- training[-inTrain, ]
clean1 <- colnames(myTraining)
clean2 <- colnames(myTraining[, -58])
myTesting <- myTesting[clean1]
testing <- testing[clean2]
for (i in 1:length(testing) ) {
for(j in 1:length(myTraining)) {
if( length( grep(names(myTraining[i]), names(testing)[j]) ) == 1) {
class(testing[j]) <- class(myTraining[i])
}
}
}
testing <- rbind(myTraining[2, -58] , testing)
testing <- testing[-1,]Prediction
Decision Trees
set.seed(123456)
modFitA1 <- rpart(classe ~ ., data=myTraining, method="class")
fancyRpartPlot(modFitA1)
predictionsA1 <- predict(modFitA1, myTesting, type = "class")
cmtree <- confusionMatrix(predictionsA1, myTesting$classe)
cmtree## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 2161 60 5 2 0
## B 51 1272 79 56 0
## C 20 176 1237 132 48
## D 0 10 27 887 81
## E 0 0 20 209 1313
##
## Overall Statistics
##
## Accuracy : 0.8756
## 95% CI : (0.8681, 0.8828)
## No Information Rate : 0.2845
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8426
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 0.9682 0.8379 0.9042 0.6897 0.9105
## Specificity 0.9881 0.9706 0.9420 0.9820 0.9642
## Pos Pred Value 0.9699 0.8724 0.7669 0.8826 0.8515
## Neg Pred Value 0.9874 0.9615 0.9790 0.9417 0.9795
## Prevalence 0.2845 0.1935 0.1744 0.1639 0.1838
## Detection Rate 0.2754 0.1621 0.1577 0.1131 0.1673
## Detection Prevalence 0.2840 0.1858 0.2056 0.1281 0.1965
## Balanced Accuracy 0.9781 0.9043 0.9231 0.8359 0.9374plot(cmtree$table, col = cmtree$byClass, main = paste("Decision Tree Confusion Matrix: Accuracy =", round(cmtree$overall['Accuracy'], 4)))
Random Forests
set.seed(123456)
modFitB1 <- randomForest(classe ~ ., data=myTraining)
predictionB1 <- predict(modFitB1, myTesting, type = "class")
cmrf <- confusionMatrix(predictionB1, myTesting$classe)
cmrf## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 2231 1 0 0 0
## B 1 1517 1 0 0
## C 0 0 1365 0 0
## D 0 0 2 1286 1
## E 0 0 0 0 1441
##
## Overall Statistics
##
## Accuracy : 0.9992
## 95% CI : (0.9983, 0.9997)
## No Information Rate : 0.2845
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.999
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 0.9996 0.9993 0.9978 1.0000 0.9993
## Specificity 0.9998 0.9997 1.0000 0.9995 1.0000
## Pos Pred Value 0.9996 0.9987 1.0000 0.9977 1.0000
## Neg Pred Value 0.9998 0.9998 0.9995 1.0000 0.9998
## Prevalence 0.2845 0.1935 0.1744 0.1639 0.1838
## Detection Rate 0.2843 0.1933 0.1740 0.1639 0.1837
## Detection Prevalence 0.2845 0.1936 0.1740 0.1643 0.1837
## Balanced Accuracy 0.9997 0.9995 0.9989 0.9998 0.9997plot(modFitB1)
plot(cmrf$table, col = cmtree$byClass, main = paste("Random Forest Confusion Matrix: Accuracy =", round(cmrf$overall['Accuracy'], 4)))
Generalized Boosted Regression
library(survival)
library(gbm)
library(splines)
library(parallel)
library(plyr)
set.seed(123456)
fitControl <- trainControl(method = "repeatedcv",
number = 5,
repeats = 1)
gbmFit1 <- train(classe ~ ., data=myTraining, method = "gbm",
trControl = fitControl,
verbose = FALSE)
gbmFinMod1 <- gbmFit1$finalModel
gbmPredTest <- predict(gbmFit1, newdata=myTesting)
gbmAccuracyTest <- confusionMatrix(gbmPredTest, myTesting$classe)
gbmAccuracyTest## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 2230 2 0 0 0
## B 2 1515 0 0 0
## C 0 1 1358 4 0
## D 0 0 10 1280 8
## E 0 0 0 2 1434
##
## Overall Statistics
##
## Accuracy : 0.9963
## 95% CI : (0.9947, 0.9975)
## No Information Rate : 0.2845
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9953
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 0.9991 0.9980 0.9927 0.9953 0.9945
## Specificity 0.9996 0.9997 0.9992 0.9973 0.9997
## Pos Pred Value 0.9991 0.9987 0.9963 0.9861 0.9986
## Neg Pred Value 0.9996 0.9995 0.9985 0.9991 0.9988
## Prevalence 0.2845 0.1935 0.1744 0.1639 0.1838
## Detection Rate 0.2842 0.1931 0.1731 0.1631 0.1828
## Detection Prevalence 0.2845 0.1933 0.1737 0.1654 0.1830
## Balanced Accuracy 0.9994 0.9989 0.9960 0.9963 0.9971plot(gbmFit1, ylim=c(0.9, 1))
Course Project Prediction Results
Random Forest gives an Accuracy in the myTesting dataset of 99.89%, which was more accurate that what I got from the Decision Trees and GBM. The expected out-of-sample error is 100%-99.89% = 0.11%. Thus I choose Random Forest to be the final predictor.
predictionB2 <- predict(modFitB1, testing, type = "class")
predictionB2## 2 31 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
## B A B A A E D B A A B C B A E E A B B B
## Levels: A B C D EWrite the results to a text file for submission
pml_write_files = function(x){
n = length(x)
for(i in 1:n){
filename = paste0("problem_id_",i,".txt")
write.table(x[i],file=filename,quote=FALSE,row.names=FALSE,col.names=FALSE)
}
}
pml_write_files(predictionB2)