library(ithir)
library(randomForest)
###################
######
DSM_data<-as.data.frame(DSM_data)
DSM_data <- na.omit(DSM_data)
str(DSM_data)
which(!complete.cases(DSM_data))
##Random Forests
library(randomForest)
set.seed(41)
training<-sample(nrow(DSM_data),0.9*nrow(DSM_data))

#fit the model
ec.RF.Exp<-randomForest(pM2.5~.,
                        
                        data=DSM_data[training,],
                        importance=T,ntree=1000,ntry=5)
#print(ec.RF.Exp)
varImpPlot(ec.RF.Exp)##MSE
#Internal validation
RF.pred.C<-predict(ec.RF.Exp,newdata=DSM_data[training,])
#RF.pred.C<-as.matrix(RF.pred.C)
goof(observed = DSM_data$pM2.5[training],predicted = RF.pred.C)

#External validation
RF.pred.V<-predict(ec.RF.Exp,newdata=DSM_data[-training,])
#RF.pred.V<-as.matrix(RF.pred.V)
goof(observed = DSM_data$pM2.5[-training],predicted = RF.pred.V)