PracticalMachineLearning.Rmd

## Download training and testing datasets
url <- "https://d396qusza40orc.cloudfront.net/predmachlearn/pml-training.csv"
download.file(url, destfile = "training.csv")
url <- "https://d396qusza40orc.cloudfront.net/predmachlearn/pml-testing.csv"
download.file(url, destfile = "testing.csv")
training <- read.csv("training.csv")
testing <- read.csv("testing.csv") 

### Clean training and testing data
training_clean <- read.csv("training.csv", na.strings = c("NA", "#DIV/0!", "")) 
testing_clean <- read.csv("testing.csv",  na.strings = c("NA", "#DIV/0!", ""))

## Remove unnecessary parameters from data
training_clean <- training_clean[, 7:160] 
testing_clean  <- testing_clean[, 7:160]  
valid_data  <- apply(!is.na(training_clean), 2, sum) > 19621  # which is the number of observations 
training_clean <- training_clean[, valid_data] 
testing_clean <- testing_clean[, valid_data] 

### Seperate out data for cross-validation
library(caret)
inTrain <- createDataPartition(y=training_clean$classe, p=0.75, list=FALSE)
train1<-training_clean[inTrain,]
train2<-training_clean[-inTrain,]

### Remove near zero variate data
nzv <- nearZeroVar(train1) 
if(length(nzv) > 0) {
 train1 <- train1[, -nzv]
 train2 <- train2[, -nzv]
}

### Build the model
myModel <- train(classe ~ ., data=train1, method="rf")

### Check the model on validation data
predTest <- predict(myModel, newdata=train2)

###Check the accuracy of the model
confusionMatrix(predTest, train2$classe)

### Predict the model on testing data
predictions<-predict(myModel, newdata=testing_clean)
testing$classe <- predictions

###Ready for outputs
testing_classe
## [1] 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 E