CODEWITHSUNDEEP
rprediction
Soly
Soly

Reputation: 121

How can i use predict() with newdata =var_i?

I'm dealing with a large data that I splitted in chunks so it's manageble by the ram, something like this: (this is an example I have more chunks)

var_1<-all_modell [c(1:150000)     ,]; save(var_1,file="~/var_1.Rdata");rm(var_1);
var_2<-all_modell [c(150001:300000),]; save(var_2,file="~/var_2.Rdata");rm(var_2);
var_3<-all_modell [c(300001:450000),]; save(var_3,file="~/var_3.Rdata");rm(var_3);

The idea is that each iteration a chunk is loaded, used to predict and then erased, so the ram is free to process the next chunk:

for (i in 1:n_chunks)
{
name<-sprintf('var_%i',i); path<-sprintf('~/var_%i.Rdata',i)
load(path)
predicted     <- predict(Model, newdata =name, type = "prob") #here is the problem
value         <- as.numeric(lapply(predicted,"[[",2))
namef         <- sprintf('~/predicted%i.Rdata',i)
save(value,file=namef)
rm(list= ls()[!(ls()%in% Model)])
}

What I would like to know is how can I pass newdata=name where name varies... I also tried this but it didn't work:

predicted <- predict(Model, parse(text=sprintf(sprintf('newdata=var_%i',i))), type="prob")

Upvotes: 1

Views: 148

Answers (2)

Greg Snow
Greg Snow

Reputation: 49640

It might be a little cleaner to use a seperate environment to hold your chunks instead of using get (though get is the simple answer here and is part of FAQ 7.21).

A possible modification of your code:

myenv <- new.env()

for (i in 1:n_chunks) {
  name<-sprintf('var_%i',i); path<-sprintf('~/var_%i.Rdata',i)
  load(path, env=myenv)
  predicted     <- predict(Model, newdata =myenv[[name]], type = "prob")
  value         <- as.numeric(lapply(predicted,"[[",2))
  namef         <- sprintf('~/predicted%i.Rdata',i)
  save(value,file=namef)
  rm(list= ls(env=myenv), envir=myenv)
}

Upvotes: 0

Andrie
Andrie

Reputation: 179428

Use get() to do this. Here is a minimal example:

x <- 1:100

x_1 <- x[1:50]
x_2 <- x[51:100]

for(i in 1:2){
  var <- sprintf('x_%i',i)
  print(sum(get(var)))
}

This results in:

[1] 1275
[1] 3775

See ?get for more details.

Upvotes: 1

Related Questions