# Customising Objective Function in R

I wondered if there were R packages that allow the user to customise the loss function?

For example, if a random forest package like ranger had a loss function which minimises OOB MSE. Is it somehow possible to customise this into a model which minimises negative log likelihood?

Would appreciate if someone knew of examples/code of doing this on a toy dataset

• I’d be shocked if you couldn’t minimize the usual loss function in classification problems. It goes by several names, “log loss” and “crossentropy loss” being two. “Negative log likelihood” comes from the fact that minimizing NLL is equivalent to maximum likelihood estimation in logistic regression.
– Dave
Commented Feb 18, 2022 at 18:25
• Hi @Dave. Valid point. I just cant see examples of being able to do it easily in R. For example, I wanted to use NLL as an out of sample comparison between models (some decision trees, transformation trees, and random forest). My baseline is a transformation forest, but it's difficult to compare performance against other models, because I would have to obtain the NLL for the other trees/RFs. Commented Feb 19, 2022 at 11:50

Using XGBoost it is relatively easy to invoke a custom loss function. There are also quite a lot of already implemented options.

It would look something like:

library(ISLR)
library(xgboost)
library(tidyverse)
library(Metrics)

# Data
df = ISLR::Hitters %>% select(Salary,AtBat,Hits,HmRun,Runs,RBI,Walks,Years,CAtBat,CHits,CHmRun,CRuns,CRBI,CWalks,PutOuts,Assists,Errors)
df = df[complete.cases(df),]
train = df[1:150,]
test = df[151:nrow(df),]

# XGBoost Matrix
dtrain <- xgb.DMatrix(data=as.matrix(train[,-1]),label=as.matrix(train[,1]))
dtest <- xgb.DMatrix(data=as.matrix(test[,-1]),label=as.matrix(test[,1]))
watchlist <- list(eval = dtest)

# Custom objective function (Huber)

myobjective <- function(preds, dtrain) {
labels <- getinfo(dtrain, "label")
d = preds - labels
h = 5
scale = 1 + (d / h)^2
scale_sqrt = sqrt(scale)
hess = 1 / scale / scale_sqrt
}

# Custom Metric
evalerror <- function(preds, dtrain) {
labels <- getinfo(dtrain, "label")
u = (preds-labels)^2
err <- (sum(u) / length(u))^(1/2)
return(list(metric = "MyError", value = err))
}

# Model Parameter
param <- list(booster = 'gbtree'
, learning_rate = 0.1
, objective = myobjective
, eval_metric = evalerror
, set.seed = 2020)

# Train Model
xgb <- xgb.train(params = param
, data = dtrain
, nrounds = 500
, watchlist
, maximize = FALSE
, early_stopping_rounds = 5
,verbose=1)

# Predict
pred = predict(xgb, dtest)
mae = mae(test\$Salary, pred)
print(mae)

• Hi @Peter. Thanks very much for your help here. Gonna digest that. I might ask some silly questions as a follow up. Commented Feb 18, 2022 at 21:41

I have done this before. Not with ranger. Packages that are open source, and with the appropriate license, I have changed the code for my model. The code might be in R, C, Python, Java.

I have also used xgboost for this. xgboost allow you to customize the objective function. You create the gradient and the hessian based on your function. You can call xgboost from R. The example is in Python. I am not sure if they support an R customized loss function but writing one function in Python might not be hard for your team.

I have not done this but you can use Tensorflow in R and write a customized loss function.