Reputation: 171
I am trying to write a function that changes the class of the variables in a dataset in order to match the class of the column of an other dataset.
By example, if I have a dataset 1 with a column named A which as a class factor, and a dataset 2 with a column named A which has a class character, I want to change the class of the column A in dataset 2 to factor.
I have a code, but I don't understand why, it fails to change the classe. Do you have any idea?
change_class2=function(predict_set,train_set){
col_drop=c()
for(column in colnames(predict_set)){
if(!column %in% colnames(train_set))
{col_drop=c(col_drop,column)}
v=grep(column, colnames(predict_set))
w=grep(column, colnames(train_set))
if((class(predict_set[,v])!=class(train_set[,w]))*(is.factor(train_set[,w]))==1){
predict_set[,v]=factor(predict_set[,v])}
else if((class(predict_set[,v])!=class(train_set[,w]))*(is.character(train_set[,w]))==1){
predict_set[,v]=as.character(predict_set[,v])}
else if((class(predict_set[,v])!=class(train_set[,w]))*(is.numeric(train_set[,w]))==1){
predict_set[,v]=as.numeric(predict_set[,v])}
else if((class(predict_set[,v])!=class(train_set[,w]))*(is.integer(train_set[,w]))==1){
predict_set[,v]=as.integer(predict_set[,v])}
else{
predict_set[,v]=predict_set[,v]}
} }
Upvotes: 1
Views: 756
Reputation: 3945
You can use the class
function to assign a class to an object:
class(predict_set[, v]) <- class(train_set[, w])
Taking it a step further, you can refer to columns as strings instead of their indices so you can take out your v
and w
grep
statements and just use column
:
class(predict_set[, column]) <- class(train_set[, column])
So all together:
change_class2 <- function(predict_set, train_set) {
for (column in colnames(predict_set)) {
class(predict_set[, column]) <- class(train_set[, column])
}
}
Upvotes: 1