Copying factors from previous data frames in R

Question

I'd like to copy the factor levels from a pre-existing data frame into a newly created data frame, rather than assigning the levels by hand.

In order to use the 'predict' function, R requires that the new data be in a data frame where the factors are the same as the that of the model-training data. I'd like to believe that the factors could be copied from the training data to the new data frame. I have gotten this to work, as shown in the code below, albeit clumsily.

# Build the model
naive_model <- NaiveBayes(outcome ~ purpose_ + home_ + emp_len_, data = loan_data, na.action = na.omit)

# Create new data point to be tested
new_loan_frame <- data.frame(purpose_ = "small_business", home_ = "MORTGAGE", emp_len_ = "> 1 Year")

# Add the necessary factors to match the training data
new_loan_frame$purpose_ <- factor(new_loan_frame$purpose_, levels = c("credit_card","debt_consolidation", "home_improvement", "major_purchase", "medical","other","small_business"))
new_loan_frame$home_ <- factor(new_loan_frame$home_, levels = c("MORTGAGE", "OWN", "RENT"))
new_loan_frame$emp_len_ <- factor(new_loan_frame$emp_len_, levels = c("< 1 Year", "> 1 Year"))

# Run the prediction using the model and the new data
predict(naive_model, new_loan_frame)

Writing out the factors for each input type seems more onerous than I'd expect should be necessary. What would be the best way to clean this up?

Answer 1

You can automate all of it.

for(cn in colnames(loan_data)) {
  new_loan_frame[,cn] <- factor(new_loan_frame[,cn], levels=levels(loan_data[,cn]))
}

Answer 2

Hi and welcome to Stackoverflow, It is correct that in order to predict, you have to have your data very well organized in a dataframe. please, try this:

new_loan_frame <- data.frame(purpose= rep(levels(loan_data$purpose),3), home = rep(levels(loan_data$home),each=7), emp_len=rep(levels(loan_data$emp_len)))

Preds1<-predict(naive_model , newdata=new_load_frame, level=0)

Additionally, try aways to not use "_" in the level names. instead, you can simply use: , sep="_")

Good luck