Classes Work Machine Larning
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.
library(caret)
library(randomForest)
library(rpart)
library(RColorBrewer)
library(e1071)Loading files
trainUrl <- "Data/pml-training.csv"
testUrl <- "Data/pml-testing.csv"
training <- read.csv(trainUrl, na.strings=c("NA","#DIV/0!",""))
testing <- read.csv(testUrl, na.strings=c("NA","#DIV/0!","")) Cleaning files
trainingset<-training[,colSums(is.na(training)) == 0]
testingset.final<-testing
testingset.final <- testingset.final[,colSums(is.na(testingset.final)) == 0]
testingset.final <- testingset.final[,-c(1:7)]
trainingset <- trainingset[,-c(1:7)]Making Partitions
inTrain <- createDataPartition(y=trainingset$classe, p=0.75, list=FALSE)
myTraining <- trainingset[inTrain, ]
myTesting <- trainingset[-inTrain, ]
dim(myTraining)## [1] 14718 53dim(myTesting)## [1] 4904 53Description of the Dataset
In this graphic I show the distribution of the classes in the dataset.
plot(myTraining$classe, col="green", main="Plot of levels of variable classe", xlab="classe", ylab="Frequency")
We Can see that there is a mayority of individuals with class A, but the rest of the dataset is very istributed between the other classes.
Model Creation
I created severeal modeles to try to fit the data and then select the best ones for the a joint or individual prediction.
Random forests
model.rf <-train(classe~., data=myTraining, method="rf", type="class")## Warning: model fit failed for Resample01: mtry= 2 Error : no se puede ubicar un vector de tamaño 112.3 Mb## Warning in nominalTrainWorkflow(x = x, y = y, wts = weights, info =
## trainInfo, : There were missing values in resampled performance measures.prediction2 <- predict(model.rf, myTesting)
confusionMatrix(prediction2, myTesting$classe)## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 1395 7 0 0 0
## B 0 941 4 0 1
## C 0 1 849 7 3
## D 0 0 2 796 2
## E 0 0 0 1 895
##
## Overall Statistics
##
## Accuracy : 0.9943
## 95% CI : (0.9918, 0.9962)
## No Information Rate : 0.2845
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9928
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 1.0000 0.9916 0.9930 0.9900 0.9933
## Specificity 0.9980 0.9987 0.9973 0.9990 0.9998
## Pos Pred Value 0.9950 0.9947 0.9872 0.9950 0.9989
## Neg Pred Value 1.0000 0.9980 0.9985 0.9981 0.9985
## Prevalence 0.2845 0.1935 0.1743 0.1639 0.1837
## Detection Rate 0.2845 0.1919 0.1731 0.1623 0.1825
## Detection Prevalence 0.2859 0.1929 0.1754 0.1631 0.1827
## Balanced Accuracy 0.9990 0.9952 0.9951 0.9945 0.9965Svm
model.svm <-svm(classe~., data=myTraining)
prediction3 <- predict(model.svm, myTesting, type = "class")
confusionMatrix(prediction3, myTesting$classe)## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 1392 67 1 1 0
## B 2 861 24 0 3
## C 1 20 818 76 29
## D 0 1 12 726 27
## E 0 0 0 1 842
##
## Overall Statistics
##
## Accuracy : 0.946
## 95% CI : (0.9393, 0.9521)
## No Information Rate : 0.2845
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9315
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 0.9978 0.9073 0.9567 0.9030 0.9345
## Specificity 0.9803 0.9927 0.9689 0.9902 0.9998
## Pos Pred Value 0.9528 0.9674 0.8665 0.9478 0.9988
## Neg Pred Value 0.9991 0.9781 0.9907 0.9812 0.9855
## Prevalence 0.2845 0.1935 0.1743 0.1639 0.1837
## Detection Rate 0.2838 0.1756 0.1668 0.1480 0.1717
## Detection Prevalence 0.2979 0.1815 0.1925 0.1562 0.1719
## Balanced Accuracy 0.9891 0.9500 0.9628 0.9466 0.9671Create a Conjunct prediction
Using Random forest and SVM
sum.pred2<-data.frame(prediction2,prediction3,classe=myTesting$classe)
mode.all2<-train(classe~.,method="rf",data=sum.pred2)
prediction.all2 <- predict(mode.all2, sum.pred2)
confusionMatrix(prediction.all2, myTesting$classe)## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 1395 7 0 0 0
## B 0 941 4 0 1
## C 0 1 849 7 3
## D 0 0 2 796 2
## E 0 0 0 1 895
##
## Overall Statistics
##
## Accuracy : 0.9943
## 95% CI : (0.9918, 0.9962)
## No Information Rate : 0.2845
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9928
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 1.0000 0.9916 0.9930 0.9900 0.9933
## Specificity 0.9980 0.9987 0.9973 0.9990 0.9998
## Pos Pred Value 0.9950 0.9947 0.9872 0.9950 0.9989
## Neg Pred Value 1.0000 0.9980 0.9985 0.9981 0.9985
## Prevalence 0.2845 0.1935 0.1743 0.1639 0.1837
## Detection Rate 0.2845 0.1919 0.1731 0.1623 0.1825
## Detection Prevalence 0.2859 0.1929 0.1754 0.1631 0.1827
## Balanced Accuracy 0.9990 0.9952 0.9951 0.9945 0.9965Predicting the test Data Set
prediction.test2<-predict(model.svm,testingset.final)
prediction.test3<-predict(model.rf,testingset.final)We Check the results for both models and if there is any difference we would need to use a rank system
prediction.test2==prediction.test3## [1] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [15] TRUE TRUE TRUE TRUE TRUE TRUEThe prediction is the same for the 2 models, so there is no need of a ranking method for them.
Testing Prediction Result
## 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
## 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