Predict Heart Disease Application

This application is intended to help doctors in the diagnosis of Coronary Heart Disease (CHD).

Using a prediction algorithm, this application indicates whether a patient could be or not diagnosed with CHD based on several risk factors.

The prediction was based on the SAHeart dataset from the ElemStatLearn packages. A retrospective sample of males in a heart-disease high-risk region of the Western Cape, South Africa. There are roughly two controls per case of CHD.

library(ElemStatLearn)
data(SAheart)
names(SAheart)

 [1] "sbp"       "tobacco"   "ldl"       "adiposity" "famhist"  
 [6] "typea"     "obesity"   "alcohol"   "age"       "chd"      

The algorithm used was prediction trees using 10-fold CV repeated 20 times. 60% of the data was used for training, 40% was used to test it.

fitControl <- trainControl(method = "repeatedcv",number = 10,repeats = 20)
rpart.grid <- expand.grid(.cp=0.005) 
train.rpart <- train(chd ~ ., data = SAheart, 
method="rpart",trControl=fitControl, tuneGrid=rpart.grid)

The algorithm was validated against the remaining test data obtaining 81% accuracy.

predSArp <- predict(train.rpart, testSA)
cmSArp <- confusionMatrix(testSA$chd, predSArp)
tab <- table(testSA$chd, predSArp)
rownames(tab)=c("Sample no CHD", "Sample CHD")
print(knitr::kable(tab, col.names=c("Predict no CHD", "Predict CHD")))

Very simple application, just enter the patient information and press the Predict button.