project.Rmd

# Practical Machine Learning Course Project



## Introduction
The aim of this project is to use data from accelerometers on the belt, forearm, arm, and dumbell of 6 participants to predict the manner in which they did the exercise. They were asked to perform barbell lifts correctly and incorrectly in 5 different ways. More information is available [here](http://groupware.les.inf.puc-rio.br/har).
 
 
 
## Data
The training and test data sets are available through the links below : <br/><br/>
  [Training Data](https://d396qusza40orc.cloudfront.net/predmachlearn/pml-training.csv) <br/>
  [Test Data](https://d396qusza40orc.cloudfront.net/predmachlearn/pml-testing.csv)


 
## Libraries
```{r eval=FALSE}
library(caret)
library(gbm)
library(randomForest)
````
 
 
 
## Loading and Cleaning Data
We start by loading the training and test data. On loading, some unknown values are transformed to NA values using the `na.strings` function.

```{r eval=FALSE}
#Load Data and convert some values to NA
training <- read.csv("pml-training.csv", na.strings=c("NA","#DIV/0!",""))
testing <- read.csv("pml-testing.csv", na.strings=c("NA","#DIV/0!",""))
````

Afterwards, we eliminate all columns containing NA values as they constitute a majority of the values in those columns as we can see from the `summary(training)` function. Columns containing irrelevant information are also stripped off. Same applies for columns containing similar values using the `nearZeroVar` function. In our case, there was no column containing values exhibiting this property.


```{r eval=FALSE}
set.seed(12479)
#Remove columns containing NA values since the majority of values in these columns are NA values
training <- training[ , sapply(training, function(x) !any(is.na(x)))]
testing <- testing[ , sapply(testing, function(x) !any(is.na(x)))]

#Remove columns containing information irrelevant to the prediction model
training <- training[ , -c(1 : 7)]
testing <- testing[ , -c(1 : 7)]

nzv <- nearZeroVar(training[, -ncol(training)])
if(length(nzv) > 0) training <- training[, -nzv]

nzv <- nearZeroVar(testing)
if(length(nzv) > 0) testing <- testing[, -nzv]
````
 
 
 
 
## Building and Training Models

* **Grading Boosting Machine**

5-fold Cross validation resampling was applied in this case:

```{r eval=FALSE}
modelGBM <- train(classe ~ ., method = "gbm", data = training, trControl = trainControl(method = "cv", number = 5), verbose = FALSE)

modelGBM
```


    Stochastic Gradient Boosting 

    19622 samples
      52 predictor
        5 classes: 'A', 'B', 'C', 'D', 'E' 

    No pre-processing
    Resampling: Cross-Validated (5 fold) 

    Summary of sample sizes: 15698, 15696, 15697, 15699, 15698 

    Resampling results across tuning parameters:

    interaction.depth  n.trees  Accuracy  Kappa  Accuracy SD  Kappa SD
    1                   50      0.752     0.685  0.01339      0.01737 
    1                  100      0.822     0.774  0.00996      0.01275 
    1                  150      0.854     0.815  0.00610      0.00777 
    2                   50      0.858     0.820  0.00888      0.01127 
    2                  100      0.907     0.882  0.00672      0.00849 
    2                  150      0.932     0.914  0.00545      0.00690 
    3                   50      0.895     0.868  0.00612      0.00776 
    3                  100      0.942     0.927  0.00573      0.00724 
    3                  150      0.962     0.952  0.00263      0.00332 

    Tuning parameter 'shrinkage' was held constant at a value of 0.1
    Accuracy was used to select the optimal model using  the largest value.
    The final values used for the model were n.trees = 150, interaction.depth = 3 and shrinkage = 0.1. 



```{r eval=FALSE}
predictions <- predict(modelGBM, newdata=training)
confusionMatrix(predictions,training$classe)         
```


    Confusion Matrix and Statistics

              Reference
    Prediction    A    B    C    D    E
            A 5528   87    0    2    4
            B   40 3636   84    8   20
            C    6   73 3297   90   27
            D    5    1   35 3101   35
            E    1    0    6   15 3521

    Overall Statistics
                                          
               Accuracy : 0.9725          
                 95% CI : (0.9701, 0.9748)
    No Information Rate : 0.2844          
    P-Value [Acc > NIR] : < 2.2e-16       
                                          
                  Kappa : 0.9652 
                  
We recorded a **97.25%** prediction accuracy.


* **Random Forest**


```{r eval=FALSE}
modelRF <- train(classe ~ ., method = "rf", data = training, trControl = trainControl(method = "cv", number = 5), verbose = FALSE)

modelRF
```



    Random Forest 

     19622 samples
      52 predictor
        5 classes: 'A', 'B', 'C', 'D', 'E' 

    No pre-processing
    Resampling: Cross-Validated (5 fold) 

    Summary of sample sizes: 15698, 15698, 15696, 15697, 15699 

    Resampling results across tuning parameters:

      mtry  Accuracy  Kappa  Accuracy SD  Kappa SD
      2    0.994     0.993  0.00201      0.00255 
      27    0.995     0.993  0.00147      0.00186 
      52    0.990     0.987  0.00326      0.00413 

    Accuracy was used to select the optimal model using  the largest value.
    The final value used for the model was mtry = 27. 
    

```{r eval=FALSE}

predictions <- predict(modelRF, newdata=training)
confusionMatrix(predictions,training$classe)        
```


    Confusion Matrix and Statistics

              Reference
    Prediction    A    B    C    D    E
            A 5580    0    0    0    0
            B    0 3797    0    0    0
            C    0    0 3422    0    0
            D    0    0    0 3216    0
            E    0    0    0    0 3607

    Overall Statistics
                                     
               Accuracy : 1          
                 95% CI : (0.9998, 1)
    No Information Rate : 0.2844     
    P-Value [Acc > NIR] : < 2.2e-16  
                                     
                  Kappa : 1  
                  
We recorded a **100%** prediction accuracy.



## Prediction and Submission

After prediction, the code provided on the course website was used to prepare files for submission.


```{r eval=FALSE}
predictions <- predict(modelRF, testing)
predictions


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(predictions)
```