Neuralnet providing different results for same predictors of different arrangements

[javakh]

I'm using the 'neuralnet' package for creating a neural network for prediction purposes. However, I'm getting different network and output results if I change the order of predictors in the same dataset. For example, say the order of predictors initially x1,x2,x3,x4,x5. But if I make it x1,x5,x3,x4,x2, I'm getting different networks and results (Shown in the attached screenshot).
I'm attaching two sets of data ( same data with different order of the predictor) and the screenshots of the different outputs I'm getting using the 'neuralnet' package. I'm also providing here the code I've used. Please check it and let me know what is wrong with it.

Code:

library(caret)
bob_post=read.csv('Set1.csv')
bob_post_scaled=as.data.frame(scale(bob_post,center=T,scale=T))
set.seed(123)
indexes = createDataPartition(bob_post$CF, p = .7, list = F)
train = bob_post_scaled[indexes, ]
test = bob_post_scaled[-indexes, ]
library(neuralnet)
set.seed(123)
model_ann <- neuralnet( CF~.,data=train , hidden = c(4,1), linear.output = T, algorithm = "rprop+")
plot(model_ann)
ann.pred_train = predict(model_ann, train)
ann.pred_test = predict(model_ann, test)
print(c(cor(ann.pred_train,train$CF),cor(ann.pred_test,test$CF)))

library(caret)
bob_post=read.csv('Set2.csv')
bob_post_scaled=as.data.frame(scale(bob_post,center=T,scale=T))
set.seed(123)
indexes = createDataPartition(bob_post$CF, p = .7, list = F)
train = bob_post_scaled[indexes, ]
test = bob_post_scaled[-indexes, ]
library(neuralnet)
set.seed(123)
model_ann <- neuralnet( CF~.,data=train , hidden = c(4,1), linear.output = T, algorithm = "rprop+")
plot(model_ann)
ann.pred_train = predict(model_ann, train)
ann.pred_test = predict(model_ann, test)
print(c(cor(ann.pred_train,train$CF),cor(ann.pred_test,test$CF)))

Set1.csv
Set2.csv

[mnwright]

Isn't that just random variation? For different column orders, weights get initialized differently which gives you different results. That should be similar to re-running with same column order but different seeds.