Import Library

library(caret)

## Loading required package: lattice

## Loading required package: ggplot2

library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

library(rattle)

## Loading required package: tibble

## Loading required package: bitops

## Rattle: A free graphical interface for data science with R.
## Version 5.4.0 Copyright (c) 2006-2020 Togaware Pty Ltd.
## Type 'rattle()' to shake, rattle, and roll your data.

library(randomForest)

## randomForest 4.6-14

## Type rfNews() to see new features/changes/bug fixes.

## 
## Attaching package: 'randomForest'

## The following object is masked from 'package:rattle':
## 
##     importance

## The following object is masked from 'package:dplyr':
## 
##     combine

## The following object is masked from 'package:ggplot2':
## 
##     margin

Download data CSV files

download.file('https://d396qusza40orc.cloudfront.net/predmachlearn/pml-training.csv',"training.csv", quiet=FALSE)
download.file('https://d396qusza40orc.cloudfront.net/predmachlearn/pml-testing.csv',"testing.csv", quiet=FALSE)

Read Data in CSV

train_data<- read.csv("training.csv")
test_data<- read.csv("testing.csv")

Cleaning the data

train_data <- read.csv('training.csv', na.strings = c("NA", "#DIV/0!", ""))
test_data <-  read.csv('testing.csv', na.strings = c("NA", "#DIV/0!", ""))
nz <- nearZeroVar(train_data)
train_data <- train_data[,-nz]
test_data <- test_data[,-nz]
rm_na <- sapply(train_data, function(x) mean(is.na(x))) > 0.95
train_data <- train_data[,rm_na == FALSE]
test_data <- test_data[,rm_na == FALSE]
train_data<- train_data[, -c(1:7)]
test_data<- test_data[, -c(1:7)]

Split the Training Dataset

inTrainIndex <- createDataPartition(train_data$classe, p=0.6)[[1]]
train_data <- train_data[inTrainIndex,]
testcv_data <- train_data[-inTrainIndex,]

Machine Learning Algorithm for Prediction

• Decision Tree
• Random Forest

dt_model <- train(classe ~., method='rpart', data=train_data)
dt_Prediction <- predict(dt_model, testcv_data)
confusionMatrix(testcv_data$classe, dt_Prediction)

## Confusion Matrix and Statistics
## 
##           Reference
## Prediction    A    B    C    D    E
##          A 1202   24   96    0    5
##          B  397  276  218   34    1
##          C  371   35  395    0    0
##          D  351   10  292  137    0
##          E  184  149  221   36  262
## 
## Overall Statistics
##                                           
##                Accuracy : 0.4838          
##                  95% CI : (0.4694, 0.4982)
##     No Information Rate : 0.5334          
##     P-Value [Acc > NIR] : 1               
##                                           
##                   Kappa : 0.3264          
##                                           
##  Mcnemar's Test P-Value : <2e-16          
## 
## Statistics by Class:
## 
##                      Class: A Class: B Class: C Class: D Class: E
## Sensitivity            0.4798  0.55870  0.32324  0.66184  0.97761
## Specificity            0.9429  0.84531  0.88313  0.85453  0.86676
## Pos Pred Value         0.9058  0.29806  0.49313  0.17342  0.30751
## Neg Pred Value         0.6132  0.94218  0.78768  0.98208  0.99844
## Prevalence             0.5334  0.10520  0.26022  0.04408  0.05707
## Detection Rate         0.2560  0.05877  0.08411  0.02917  0.05579
## Detection Prevalence   0.2826  0.19719  0.17057  0.16823  0.18143
## Balanced Accuracy      0.7114  0.70201  0.60319  0.75818  0.92218

rf_model <- train(classe ~., method='rf', data=train_data, ntree=10)
rf_prediction <- predict(rf_model, testcv_data)
confusionMatrix(testcv_data$classe, rf_prediction)

## Confusion Matrix and Statistics
## 
##           Reference
## Prediction    A    B    C    D    E
##          A 1327    0    0    0    0
##          B    0  926    0    0    0
##          C    0    1  800    0    0
##          D    0    0    1  789    0
##          E    0    0    0    0  852
## 
## Overall Statistics
##                                           
##                Accuracy : 0.9996          
##                  95% CI : (0.9985, 0.9999)
##     No Information Rate : 0.2826          
##     P-Value [Acc > NIR] : < 2.2e-16       
##                                           
##                   Kappa : 0.9995          
##                                           
##  Mcnemar's Test P-Value : NA              
## 
## Statistics by Class:
## 
##                      Class: A Class: B Class: C Class: D Class: E
## Sensitivity            1.0000   0.9989   0.9988   1.0000   1.0000
## Specificity            1.0000   1.0000   0.9997   0.9997   1.0000
## Pos Pred Value         1.0000   1.0000   0.9988   0.9987   1.0000
## Neg Pred Value         1.0000   0.9997   0.9997   1.0000   1.0000
## Prevalence             0.2826   0.1974   0.1706   0.1680   0.1814
## Detection Rate         0.2826   0.1972   0.1704   0.1680   0.1814
## Detection Prevalence   0.2826   0.1972   0.1706   0.1682   0.1814
## Balanced Accuracy      1.0000   0.9995   0.9992   0.9999   1.0000

Result

• From the confusion matrix it is clear that random forest algorithm works better than decision tree. So using random forest model the prediction should be made.

Conclusion

predict(rf_model, test_data)

##  [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