added solution for classification

R/BIA_T09_Classification.Rmd

@@ -157,4 +157,62 @@ selected.features <- cfs(formula=as.simple.formula(class="pNumResidents", attrib
 #further feature filter
 selected.features2 <- consistency(formula=as.simple.formula(class="pNumResidents", 
                                                             attributes = all.features), data = alldata)
-```
\ No newline at end of file
+```
+
+
+```{r Classification Basic evaluation approach}
+set.seed(1506)
+test.cases <- match(
+  sample(alldata$VID, size = .3*nrow(alldata)), 
+  alldata$VID)
+
+## decisoon tree
+
+#train the model
+model <- ctree(pNumResidents ~ . , 
+               data=alldata[-test.cases, c("pNumResidents", selected.features)])
+
+#predict test cases
+clres <- predict(model, newdata=alldata[test.cases,c("pNumResidents", selected.features)])
+
+#create confusion matrix and calculate accuracy
+cm <- table(clres, alldata$pNumResidents[test.cases])
+(accuracy <- (sum(diag(cm))/sum(as.vector(cm))))
+
+## random forest
+
+#train the model
+model <- randomForest(pNumResidents ~ . , 
+                 data=alldata[-test.cases, c("pNumResidents", selected.features)])
+
+#predict test cases
+clres <- predict(model, newdata=alldata[test.cases,c("pNumResidents", selected.features)])
+
+#create confusion matrix and calculate accuracy
+cm <- table(clres, alldata$pNumResidents[test.cases])
+(accuracy <- (sum(diag(cm))/sum(as.vector(cm))))
+
+## kNN
+testdata <- na.omit(alldata[test.cases, c("pNumResidents", selected.features)])
+traindata <- na.omit(alldata[-test.cases, c("pNumResidents", selected.features)])
+
+# predict test cases from training data (lazy learning algoritm has no explicit training step!)
+clres <- knn(train = traindata[,-1], test = testdata[,-1], 
+             traindata$pNumResidents, k = 5)
+
+#create confusion matrix and calculate accuracy
+cm <- table(clres, testdata$pNumResidents)
+(accuracy <- (sum(diag(cm))/sum(as.vector(cm))))
+
+## SVM
+
+#train the model
+model <- svm(pNumResidents ~ . , data=traindata)
+
+#predict the test cases
+clres <- predict(model, newdata=testdata)
+
+#create confusion matrix and calculate accuracy
+cm <- table(clres, testdata$pNumResidents)
+(accuracy <- (sum(diag(cm))/sum(as.vector(cm))))
+```