Random Forests-Based Operational Status Perception Model in Extra-Long Highway Tunnels with Longitudinal Ventilation: A Case Study in China

Algorithm 1

# division of training set and test set

set.seed(100)

ind <- sample(2, nrow(Mydataset), replace = TRUE, prob = c(0.7,0.3))

trainingset <- Mydatasetind==1, # training set accounts for 70%

testset <- Mydatasetind==2, # test set accounts for 30%

# combined parameters in Random Forests

library(randomForest)

library(caret)

M <- ncol(trainingset)

ntree <- 10c(1:50)

result <- data.frame()

# model training

set.seed(100)

for(m in 1:(M-1))

for(n in ntree)

fit.rf <- randomForest(Status ~., data = trainingset, mtry = m, ntree = n, na.action = na.omit)

OOB.ER <- 1-confusionMatrix(as.table(fit.rf$confusion[,c(-4)]))overall[Accuracy]

result <- rbind(result, data.frame(m, n, OOB.ER))

print(result)