Tuned XGBOOST model Error - BIOMOD version 4.2-5

[suicmezburak]

Hello, As the title suggests, after tuning the settings for the XGBoost model, the model is not working and is giving an error.

The sequence of code before encountering an error;

# Define individual models options
opt.d <- bm_ModelingOptions(
  data.type = "binary",
  models = c("XGBOOST"),
  strategy = "bigboss",
  bm.format = x_data,
  calib.lines = cv.k
  )

### tune parameters for XGBOOST
tuned.xgboost <- bm_Tuning(
  model = "XGBOOST", 
  tuning.fun = "xgbTree", 
  bm.options = opt.d@options$XGBOOST.binary.xgboost.xgboost,
  bm.format = x_data,
  calib.lines = cv.k
)

## Integrate tuned list to opt.d
user.val <- list()
user.val$XGBOOST.binary.xgboost.xgboost <- tuned.xgboost
# +Actually, I will add other models, but to keep the error code short, I removed them.

opt.d <- bm_ModelingOptions(
  data.type = "binary", 
  models = c("XGBOOST"), 
  strategy = "user.defined", 
  user.val = user.val, 
  bm.format = x_data, 
  calib.lines = cv.k
)

## Modelling
# by algorithm
x_model_output <- 
  BIOMOD_Modeling(
    bm.format = x_data,
    modeling.id = "AllModels",
    models = c("XGBOOST"),
    CV.strategy = "user.defined",
    CV.user.table = cv.k,
    CV.do.full.models = FALSE,
    OPT.user = opt.d,
    metric.eval = c("TSS", "ROC"),
    var.import = 3
  )

The error message:

-=-=-=--=-=-=- x_PA1_RUN1_XGBOOST 
Error in check.booster.params(params, ...) : 
  Same parameters in 'params' and in the call are not allowed. Please check your 'params' list.

*** Error in  XGBOOST , calibration failed
*** inherits(g.pred,'try-error')
   ! Note :  x_PA1_RUN1_XGBOOST failed!

-=-=-=--=-=-=- x_PA1_RUN2_XGBOOST 
Error in check.booster.params(params, ...) : 
  Same parameters in 'params' and in the call are not allowed. Please check your 'params' list.

*** Error in  XGBOOST , calibration failed
*** inherits(g.pred,'try-error')
   ! Note :  x_PA1_RUN2_XGBOOST failed!

-=-=-=--=-=-=- x_PA1_RUN3_XGBOOST 
Error in check.booster.params(params, ...) : 
  Same parameters in 'params' and in the call are not allowed. Please check your 'params' list.

*** Error in  XGBOOST , calibration failed
*** inherits(g.pred,'try-error')
   ! Note :  x_PA1_RUN3_XGBOOST failed!

### This error message persists for all PA and RUNs.

! All models failed

Based on the error message, as an example, I am sharing the configuration content for PA1_RUN1.

> opt.d@options[["XGBOOST.binary.xgboost.xgboost"]]@args.values[["_PA1_RUN1"]]
$missing
[1] NA

$params
list(nrounds = 50, max_depth = 1, eta = 0.4, gamma = 0, colsample_bytree = 0.6, 
    min_child_weight = 1, subsample = 0.5)

$nrounds
[1] 4

$verbose
[1] 1

$print_every_n
[1] 1

$save_name
[1] "xgboost.model"

$callbacks
list()

$max.depth
[1] 2

$eta
[1] 1

$nthread
[1] 2

$objective
[1] "binary:logistic"

The error indicates that there are duplicate parameters for XGBOOST in the settings. Is the reason for this that I created opt.d again after tuning the settings? The code was actually working, and I started getting errors without making any changes.

Thanks in advance, Burak

[HeleneBlt]

Hi !

Thanks for testing and your reproducible example. It seems some changes in Xgboost package went faster than expected.

I push some modifications into the GitHub version. Could you reinstall biomod2 and try again? 🙏

Thanks! Hélène

[suicmezburak]

indeed code worked this time, thanks!!