search

NorskRegnesentral / shapr

Explaining the output of machine learning models with more accurately estimated Shapley values
https://norskregnesentral.github.io/shapr/
Other
141 stars 33 forks source link

shapr for lightGBM #331

Closed giacomog99 closed 1 year ago

giacomog99 commented 1 year ago

Hi! I would like to know If the shapr explanation package can work with lighGBM model. I created this model:

data preparation

train_id <- createDataPartition(bank$deposit, p = .8, list = FALSE)

train = bank[train_id, ] test = bank[-train_id, ]

train and test matrix

trainm = sparse.model.matrix(deposit ~., data = train) train_label = train[,"deposit"]

testm = sparse.model.matrix(deposit~., data= test) test_label = test[,"deposit"]

train_matrix = lgb.Dataset(data = as.matrix(trainm), label = train_label) test_matrix = lgb.Dataset(data = as.matrix(testm), label = test_label)

model parameters

params = list(learning_rate = 0.01, objective = "binary")

model training

model = lightgbm(params = params, data = train_matrix, nrounds = 1000)

and now I want to use shapr explanation that I've already used for XGBoost but I'm not sure if I can use also for lightGBM

Thank you!

martinju commented 1 year ago

Hi

Yes, you can. You just need to make the prediction function yourself, see the vignette: https://norskregnesentral.github.io/shapr/articles/understanding_shapr.html#explain-custom-models