I am a newbie to neural networks and been trying out the algorithm with some big data sets. Is there a way to include all the input variables into the network without having to type all the names? For example I have around 30 variables which I'd like to use as input to predict the output. Is there a shortcut for the following command?
net <- neuralnet(Output~Var1+Var2+Var3+Var4+.....upto Var30, data, hidden=0)
I've looked everywhere but couldn't find the solution to this. Sorry if this is a basic question!
There are 3 ways to insert variables in the formula part of a function:
First by using . which will include all of the variables in the data data.frame apart from the response variable (variable Output in this case):
net <- neuralnet(Output ~ ., data, hidden=0) #apart from Output all of the other variables in data are included
Use this if your data.frame has Output and another 30 variables only.
Second if you want to use a vector of names to include from the data data.frame you can try:
names <- c('var1','var2','var3') #choose the names you want
a <- as.formula(paste('Output ~ ' ,paste(names,collapse='+')))
> a
Output ~ var1 + var2 + var3 #this is what goes in the neuralnet function below
so you can use:
net <- neuralnet( a , data, hidden=0) #use a in the function
Use this if you can provide a vector of the names of the 30 variables
Third just subset the data data.frame using the columns you want in the function e.g.:
net <- neuralnet(Output ~ ., data=data[,1:31] , hidden=0)
Use this to (or any other subset that is convenient) and choose the 30 variables you need along with the Output variable. Then use . to include everything.
Hope it helps!
来源：https://stackoverflow.com/questions/27779667/neural-network-using-all-input-variables