Nesse arquivo foi usada a seguinte estratégia:

Utilizei um recurso chamado DTW (Dynamic Time Warping) onde se calcula a dissimilaridades entre as séries temporais. Nesse caso entre cada movimento ou path realizado pelo usuário. Dessa forma podemos contabilizar o quão similar as series são entre si.

Com essa estratégia podemos calcular uma distancia e assim realizar procedimentos como clusterizações e até classificações. Nesse caso utilizei um simples KNN(k-nearest neighbors) para realizar a classificação binária.

Importante dizer que esse método (DTW) é robusto para séries com múltiplas variaveis (multivariate).

Obs. Podemos melhorar essa abordagem adicionando um filtro Kernel na contrução de um modelo KNN mais robusto.



In [20]:

    
library(dtw)
library(KODAMA)
library(caret)
library(pROC)

groups <- read.csv(file="./MovementAAL/groups/MovementAAL_DatasetGroup.csv",head=TRUE,sep=",")
targetAll <- read.csv(file="./MovementAAL/dataset/MovementAAL_target.csv",head=TRUE,sep=",")



In [21]:

    
#Group 1
allDataGroup1<-list()
allDataGroup1Target<-list()
groups1 = groups[groups$dataset_ID==1, ]

index<-1
for (id in groups1$X.sequence_ID){
    caminho <-paste("./MovementAAL/dataset/MovementAAL_RSS_",id,".csv",sep="")
    allDataGroup1[[index]]<-read.csv(file=caminho,head=TRUE,sep=",")
    allDataGroup1Target[index]<-targetAll[[2]][id]
    index<-index+1
}

kdist <- dist(allDataGroup1, method="DTW",diag = TRUE,upper = TRUE)
kdistM<- as.matrix(kdist)



In [22]:

    
target <- unlist(allDataGroup1Target)
allDataGroup1TargetFrame = as.data.frame(target)
train <- createDataPartition(allDataGroup1TargetFrame$target, p = .7, list = TRUE,times=10)
y <- allDataGroup1TargetFrame$target

Média e Desvio padrão respectivamente.

Group 1, com Cross Validation tipo 10-fold



In [23]:

    
allAccuracyGroup1 <- c()

for( i in 1:length(train)){
    test  <- (1:nrow(allDataGroup1TargetFrame))[-train[[i]]]
    preds <- knn.predict(train[[i]], test,y, kdistM, k=3,agg.meth="majority") 
    matrix <- confusionMatrix(preds, y[test])
    #accuracy
    allAccuracyGroup1 <- c(allAccuracyGroup1,matrix[3]$overall[[1]])
}
mean(allAccuracyGroup1)
sd(allAccuracyGroup1)









    Out[23]:




0.946666666666667






    Out[23]:




0.0281091347570523



In [24]:

    
#Group 2


allDataGroup2<-list()
groups2 = groups[groups$dataset_ID==2, ]
allDataGroup2Target<-list()
index<-1
for (id in groups2$X.sequence_ID){
    caminho <-paste("./MovementAAL/dataset/MovementAAL_RSS_",id,".csv",sep="")
    allDataGroup2[[index]]<-read.csv(file=caminho,head=TRUE,sep=",")
    allDataGroup2Target[index]<-targetAll[[2]][id]
    index<-index+1
}


kdist <- dist(allDataGroup2, method="DTW",diag = TRUE,upper = TRUE)
kdistM<- as.matrix(kdist)



In [25]:

    
target <- unlist(allDataGroup2Target)
allDataGroup2TargetFrame = as.data.frame(target)
train <- createDataPartition(allDataGroup2TargetFrame$target, p = .7, list = TRUE,times=10)
y <- allDataGroup2TargetFrame$target

Média e Desvio padrão respectivamente.

Group 2, com Cross Validation tipo 10-fold



In [26]:

    
allAccuracyGroup2 <- c()

for( i in 1:length(train)){
    test  <- (1:nrow(allDataGroup2TargetFrame))[-train[[i]]]
    preds <- knn.predict(train[[i]], test,y, kdistM, k=3,agg.meth="majority") 
    matrix <- confusionMatrix(preds, y[test])
    #accuracy
    allAccuracyGroup2 <- c(allAccuracyGroup2,matrix[3]$overall[[1]])
}

mean(allAccuracyGroup2)
sd(allAccuracyGroup2)









    Out[26]:




0.9






    Out[26]:




0.0635228193021158



In [27]:

    
#Group 3


allDataGroup3<-list()
groups3 = groups[groups$dataset_ID==3, ]
allDataGroup3Target<-list()
index<-1
for (id in groups3$X.sequence_ID){
    caminho <-paste("./MovementAAL/dataset/MovementAAL_RSS_",id,".csv",sep="")
    allDataGroup3[[index]]<-read.csv(file=caminho,head=TRUE,sep=",")
    allDataGroup3Target[index]<-targetAll[[2]][id]
    index<-index+1
}


kdist <- dist(allDataGroup3, method="DTW",diag = TRUE,upper = TRUE)
kdistM<- as.matrix(kdist)



In [28]:

    
target <- unlist(allDataGroup3Target)
allDataGroup3TargetFrame = as.data.frame(target)
train <- createDataPartition(allDataGroup3TargetFrame$target, p = .7, list = TRUE,times=10)
y <- allDataGroup3TargetFrame$target

Média e Desvio padrão respectivamente.

Group 3, com Cross Validation tipo 10-fold



In [29]:

    
allAccuracyGroup3 <- c()

for( i in 1:length(train)){
    test  <- (1:nrow(allDataGroup3TargetFrame))[-train[[i]]]
    preds <- knn.predict(train[[i]], test,y, kdistM, k=3,agg.meth="majority") 
    matrix <- confusionMatrix(preds, y[test])
    #accuracy
    allAccuracyGroup3 <- c(allAccuracyGroup3,matrix[3]$overall[[1]])
}

mean(allAccuracyGroup3)
sd(allAccuracyGroup3)









    Out[29]:




0.893548387096774






    Out[29]:




0.0482075521966879



In [30]:

    
#All DataBases

targetAll <- read.csv(file="./MovementAAL/dataset/MovementAAL_target.csv",head=TRUE,sep=",")
allData<-list()
allDataTarget<-list()
index<-1
for (id in targetAll$X.sequence_ID){
    caminho <-paste("./MovementAAL/dataset/MovementAAL_RSS_",id,".csv",sep="")
    allData[[index]]<-read.csv(file=caminho,head=TRUE,sep=",")
    allDataTarget[index]<-targetAll[[2]][id]
    index<-index+1
}

kdist <- dist(allData, method="DTW",diag = TRUE,upper = TRUE)
kdistM<- as.matrix(kdist)



In [31]:

    
target <- unlist(allDataTarget)
allDataTargetFrame = as.data.frame(target)
train <- createDataPartition(allDataTargetFrame$target, p = .7, list = TRUE,times=10)
y <- allDataTargetFrame$target

Média e Desvio padrão respectivamente.

Todos os Groups em uma base apenas, com Cross Validation tipo 10-fold



In [32]:

    
allAccuracy <- c()

for( i in 1:length(train)){
    test  <- (1:nrow(allDataTargetFrame))[-train[[i]]]
    preds <- knn.predict(train[[i]], test,y, kdistM, k=3,agg.meth="majority") 
    matrix <- confusionMatrix(preds, y[test])
    #accuracy
    allAccuracy <- c(allAccuracyGroup3,matrix[3]$overall[[1]])
}

mean(allAccuracy)
sd(allAccuracy)









    Out[32]:




0.892587508579272






    Out[32]:




0.0458446010201432

Matrix de confusão

Todos os Groups em uma base apenas



In [35]:

    
#All groups datasets Confusion Matrix 
target <- unlist(allDataTarget)
allDataTargetFrame = as.data.frame(target)
train <- createDataPartition(allDataTargetFrame$target, p = .7, list = TRUE,times=1)
y <- allDataTargetFrame$target
y <- factor(unlist(y),labels=c("No", "Yes"))
test  <- (1:nrow(allDataTargetFrame))[-train[[1]]]
preds <- knn.predict(train[[1]], test,y, kdistM, k=3,agg.meth="majority") 
classes <- factor(unlist( y[test]),labels=c("No", "Yes"))
matrix <- confusionMatrix(preds,classes)
matrix









    Out[35]:





Confusion Matrix and Statistics

          Reference
Prediction No Yes
       No  48   5
       Yes  4  37
                                         
               Accuracy : 0.9043         
                 95% CI : (0.826, 0.9553)
    No Information Rate : 0.5532         
    P-Value [Acc > NIR] : 1.206e-13      
                                         
                  Kappa : 0.8059         
 Mcnemar's Test P-Value : 1              
                                         
            Sensitivity : 0.9231         
            Specificity : 0.8810         
         Pos Pred Value : 0.9057         
         Neg Pred Value : 0.9024         
             Prevalence : 0.5532         
         Detection Rate : 0.5106         
   Detection Prevalence : 0.5638         
      Balanced Accuracy : 0.9020         
                                         
       'Positive' Class : No

Curva ROC e AUC

Todos os Groups em uma base apenas



In [36]:

    
#ROC CURVE AND AUC
predsProb <- t(knn.probability(train[[1]], test,y, kdistM, k=3))
outcome<- predsProb[,1]
rocobj <- roc(classes, outcome,levels=c("No", "Yes"))
plot(rocobj)









    Out[36]:





Call:
roc.default(response = classes, predictor = outcome, levels = c("No",     "Yes"))

Data: outcome in 52 controls (classes No) > 42 cases (classes Yes).
Area under the curve: 0.9423



In [ ]: