Health Data Analysis
Saif Ul Mehdi
08/07/2020
Overview
This document is the final report of the Peer Assignment from the course Practical Machine Learning offered by JHU. It was built in RStudio using knitr function and published as html document.
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
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.
The following statements were made by the original authors of the dataset.
Six young health participants were asked to perform one set of 10 repetitions of the Unilateral Dumbbell Biceps Curl in five different fashions: exactly according to the specification (Class A), throwing the elbows to the front (Class B), lifting the dumbbell only halfway (Class C), lowering the dumbbell only halfway (Class D) and throwing the hips to the front (Class E).
Class A corresponds to the specified execution of the exercise, while the other 4 classes correspond to common mistakes. Participants were supervised by an experienced weight lifter to make sure the execution complied to the manner they were supposed to simulate. The exercises were performed by six male participants aged between 20-28 years, with little weight lifting experience. We made sure that all participants could easily simulate the mistakes in a safe and controlled manner by using a relatively light dumbbell (1.25kg).
Source : Velloso, E.; Bulling, A.; Gellersen, H.; Ugulino, W.; Fuks, H. Qualitative Activity Recognition of Weight Lifting Exercises. Proceedings of 4th Augmented Human (AH) International Conference in cooperation with ACM SIGCHI (Augmented Human’13) . Stuttgart, Germany: ACM SIGCHI, 2013.
Loading Required Libraries
suppressPackageStartupMessages(library(caret))
suppressPackageStartupMessages(library(rpart))
suppressPackageStartupMessages(library(rpart.plot))
suppressPackageStartupMessages(library(rattle))
suppressPackageStartupMessages(library(randomForest))
suppressPackageStartupMessages(library(corrplot)) Loading Dataset and cleaning
We will be using the training data for analysis and testing data for answering quiz questions. Training data will be split into training and testing set in the ratio 70:30. Many columns have NAs which would be removed and the identity columns along with ID columns will be removed.
download.file('https://d396qusza40orc.cloudfront.net/predmachlearn/pml-training.csv',destfile='./training_set.csv',method='curl')
download.file('https://d396qusza40orc.cloudfront.net/predmachlearn/pml-testing.csv',destfile='./testing_set.csv',method='curl')
train_dat<-read.csv('./training_set.csv')
test_dat<-read.csv('./testing_set.csv')
inTrain<-createDataPartition(y=train_dat$classe,p=0.7,list=FALSE)
train_set<-train_dat[inTrain,]
test_set<-train_dat[-inTrain,]
NZV<-nearZeroVar(train_set)
train_set<-train_set[,-NZV]
test_set<-test_set[,-NZV]
All_NA<-sapply(train_set,function(x) mean(is.na(x)))
train_set<-train_set[,All_NA==FALSE]
test_set<-test_set[,All_NA==FALSE]
train_set<-train_set[,-(1:5)]
test_set<-test_set[,-(1:5)]After cleaning cleaning process, both training and testing sets have 54 columns
Corelation
Before creating prediction models, correlation is found among the variables
The highly correlated variables are depicted in dark colors in the above plot. Since, there are very few variables which are correlated strongly, PCA is not necessary to be performed as pre-processing step.
Prediction Model Building
Three methods are used for model building and the one with highest accuracy score will be applied to the testing data.
Random Forest
set.seed(12345)
controlRF<-trainControl(method='cv',number=3,verboseIter=FALSE)
mdlRF<-train(classe~.,data=train_set,method='rf',trControl=controlRF)
mdlRF$finalModel
Call:
randomForest(x = x, y = y, mtry = param$mtry)
Type of random forest: classification
Number of trees: 500
No. of variables tried at each split: 27
OOB estimate of error rate: 0.23%
Confusion matrix:
A B C D E class.error
A 3904 1 0 0 1 0.0005120328
B 8 2647 2 1 0 0.0041384500
C 0 5 2391 0 0 0.0020868114
D 0 0 7 2244 1 0.0035523979
E 0 0 0 5 2520 0.0019801980predRF<-predict(mdlRF,newdata=test_set)
conmatRF<-confusionMatrix(predRF,as.factor(test_set$classe))
conmatRFConfusion Matrix and Statistics
Reference
Prediction A B C D E
A 1673 3 0 0 0
B 1 1132 3 0 0
C 0 4 1023 2 0
D 0 0 0 961 3
E 0 0 0 1 1079
Overall Statistics
Accuracy : 0.9971
95% CI : (0.9954, 0.9983)
No Information Rate : 0.2845
P-Value [Acc > NIR] : < 2.2e-16
Kappa : 0.9963
Mcnemar's Test P-Value : NA
Statistics by Class:
Class: A Class: B Class: C Class: D Class: E
Sensitivity 0.9994 0.9939 0.9971 0.9969 0.9972
Specificity 0.9993 0.9992 0.9988 0.9994 0.9998
Pos Pred Value 0.9982 0.9965 0.9942 0.9969 0.9991
Neg Pred Value 0.9998 0.9985 0.9994 0.9994 0.9994
Prevalence 0.2845 0.1935 0.1743 0.1638 0.1839
Detection Rate 0.2843 0.1924 0.1738 0.1633 0.1833
Detection Prevalence 0.2848 0.1930 0.1749 0.1638 0.1835
Balanced Accuracy 0.9993 0.9965 0.9979 0.9981 0.9985
Decision Trees
set.seed(12345)
mdldtree<-rpart(classe~.,data=train_set,method='class')
fancyRpartPlot(mdldtree)## Warning: labs do not fit even at cex 0.15, there may be some overplotting
preddtree<-predict(mdldtree,newdata=test_set,type='class')
conmatdtree<-confusionMatrix(preddtree,as.factor(test_set$classe))
conmatdtreeConfusion Matrix and Statistics
Reference
Prediction A B C D E
A 1555 180 19 41 8
B 34 773 85 96 41
C 1 53 821 38 4
D 70 82 92 712 84
E 14 51 9 77 945
Overall Statistics
Accuracy : 0.8167
95% CI : (0.8065, 0.8265)
No Information Rate : 0.2845
P-Value [Acc > NIR] : < 2.2e-16
Kappa : 0.7675
Mcnemar's Test P-Value : < 2.2e-16
Statistics by Class:
Class: A Class: B Class: C Class: D Class: E
Sensitivity 0.9289 0.6787 0.8002 0.7386 0.8734
Specificity 0.9411 0.9461 0.9802 0.9333 0.9686
Pos Pred Value 0.8625 0.7512 0.8953 0.6846 0.8622
Neg Pred Value 0.9708 0.9246 0.9587 0.9480 0.9714
Prevalence 0.2845 0.1935 0.1743 0.1638 0.1839
Detection Rate 0.2642 0.1314 0.1395 0.1210 0.1606
Detection Prevalence 0.3064 0.1749 0.1558 0.1767 0.1862
Balanced Accuracy 0.9350 0.8124 0.8902 0.8360 0.9210
Generalized Boosting Method
set.seed(12345)
controlGBM<-trainControl(method='repeatedcv',number=5,repeats=1)
mdlGBM<-train(classe~.,data=train_set,method='gbm',trControl=controlGBM,verbose=FALSE)
mdlGBM$finalModelA gradient boosted model with multinomial loss function.
150 iterations were performed.
There were 53 predictors of which 53 had non-zero influence.predGBM<-predict(mdlGBM,newdata=test_set)
conmatGBM<-confusionMatrix(predGBM,as.factor(test_set$classe))
conmatGBMConfusion Matrix and Statistics
Reference
Prediction A B C D E
A 1673 13 0 0 1
B 1 1111 13 3 4
C 0 14 1010 13 2
D 0 1 2 947 6
E 0 0 1 1 1069
Overall Statistics
Accuracy : 0.9873
95% CI : (0.9841, 0.99)
No Information Rate : 0.2845
P-Value [Acc > NIR] : < 2.2e-16
Kappa : 0.9839
Mcnemar's Test P-Value : NA
Statistics by Class:
Class: A Class: B Class: C Class: D Class: E
Sensitivity 0.9994 0.9754 0.9844 0.9824 0.9880
Specificity 0.9967 0.9956 0.9940 0.9982 0.9996
Pos Pred Value 0.9917 0.9814 0.9721 0.9906 0.9981
Neg Pred Value 0.9998 0.9941 0.9967 0.9966 0.9973
Prevalence 0.2845 0.1935 0.1743 0.1638 0.1839
Detection Rate 0.2843 0.1888 0.1716 0.1609 0.1816
Detection Prevalence 0.2867 0.1924 0.1766 0.1624 0.1820
Balanced Accuracy 0.9980 0.9855 0.9892 0.9903 0.9938
Applying Best model to Test Data
The accuracy of the 3 regression models are - a) Random Forest : 0.9976 b) Decision Trees : 0.7443 c) Generalized Boosting Method : 0.9891
So, the best model is random forest.
predictTest<-predict(mdlRF,newdata=test_dat)
predictTest [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