Attribute Information:
PimaIndiansDiabetes.pca <- princomp(PimaIndiansDiabetes[,1:(length(PimaIndiansDiabetes)-1)],
center = TRUE,
scale. = TRUE)
View(PimaIndiansDiabetes.pca$scores)
target.numeric <- ifelse (PimaIndiansDiabetes$diabetes == "neg", 0 ,1)
pca.sisense.result <- data.frame(PimaIndiansDiabetes.pca$scores[,1:2],target.numeric)
x <- pca.sisense.result$Comp.1
y <- pca.sisense.result$Comp.2
pca.sisense.result$class <- as.factor(pca.sisense.result$target.numeric)
ggplot(pca.sisense.result, aes(Comp.1, Comp.2)) +
geom_point(aes(colour = class)) 
ct <- ctree(diabetes ~ ., data = training)
svg(filename="Diabites_Decision_Tree.svg",
width=12,
height=10,
pointsize=12)
plot(as.simpleparty(ct))
dev.off()## png
## 2## Predict Diabites Risk on new patients
predictions.probs <- predict(ct, testing,type = c("prob"))
predictions.class <- predict(ct, testing,type = c("response"))
table(predictions.class, testing$diabetes )##
## predictions.class neg pos
## neg 130 44
## pos 20 36## confusion matrix stats on testing set
cm <- confusionMatrix(testing$diabetes, predictions.class, positive = NULL,
dnn = c("Prediction", "Reference"))
# printing confusion matrix
cm## Confusion Matrix and Statistics
##
## Reference
## Prediction neg pos
## neg 130 20
## pos 44 36
##
## Accuracy : 0.7217
## 95% CI : (0.659, 0.7786)
## No Information Rate : 0.7565
## P-Value [Acc > NIR] : 0.90280
##
## Kappa : 0.3405
## Mcnemar's Test P-Value : 0.00404
##
## Sensitivity : 0.7471
## Specificity : 0.6429
## Pos Pred Value : 0.8667
## Neg Pred Value : 0.4500
## Prevalence : 0.7565
## Detection Rate : 0.5652
## Detection Prevalence : 0.6522
## Balanced Accuracy : 0.6950
##
## 'Positive' Class : neg
## roc_pred <- prediction(predictions.probs[,2], testing$diabetes)
#png(filename=paste("decision_tree_roc.png"),width = 1200, height = 800)
plot(performance(roc_pred, measure="sens", x.measure="spec"), colorize=TRUE)
auc.perf = performance(roc_pred, measure = "auc")
title(unlist(auc.perf@y.values))
#dev.off()
fit.rf <-randomForest(diabetes ~ ., data=training)
print(fit.rf)##
## Call:
## randomForest(formula = diabetes ~ ., data = training)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 2
##
## OOB estimate of error rate: 21.19%
## Confusion matrix:
## neg pos class.error
## neg 305 45 0.1285714
## pos 69 119 0.3670213#importance(fit.rf)
rf.predictions.class <- predict(fit.rf, testing)
rf.predictions.probs <- predict(fit.rf, testing,type = c("prob"))
cm.rf <- confusionMatrix(testing$diabetes, rf.predictions.class, positive = NULL,
dnn = c("Prediction", "Reference"))
# printing confusion matrix
cm.rf## Confusion Matrix and Statistics
##
## Reference
## Prediction neg pos
## neg 129 21
## pos 40 40
##
## Accuracy : 0.7348
## 95% CI : (0.6728, 0.7906)
## No Information Rate : 0.7348
## P-Value [Acc > NIR] : 0.53439
##
## Kappa : 0.3811
## Mcnemar's Test P-Value : 0.02119
##
## Sensitivity : 0.7633
## Specificity : 0.6557
## Pos Pred Value : 0.8600
## Neg Pred Value : 0.5000
## Prevalence : 0.7348
## Detection Rate : 0.5609
## Detection Prevalence : 0.6522
## Balanced Accuracy : 0.7095
##
## 'Positive' Class : neg
## roc_pred.rf <- prediction(rf.predictions.probs[,2], testing$diabetes)
plot(performance(roc_pred.rf, measure="sens", x.measure="spec"), colorize=TRUE)
rf.auc.perf = performance(roc_pred.rf, measure = "auc")
title(unlist(rf.auc.perf@y.values))
- an AUC improvement of
unlist(rf.auc.perf@y.values) - unlist(auc.perf@y.values)## [1] 0.03920833NOT BAD !