be-marc
commented
3 months ago Please write a minimal reproducible example. I cannot check without the data.
Closed zrq1209 closed 3 months ago
be-marc
commented
3 months ago Please write a minimal reproducible example. I cannot check without the data.
data = read_excel("F:/R data/test1.xlsx") data = as.data.table(data) class(data$biomass)#检查读取数据后的类型
target = "biomass" features = setdiff(names(data), target)
task=TaskRegr$new(data, target = "biomass",id = "prediction") task
filter = flt("correlation") filter$calculate(task) top_features = as.data.table(filter)$feature[1:10]
scores_dt = data.table(feature = names(filter$scores), score = filter$scores)
task_selected = task$select(top_features) new_data = task_selected$data() task_selected = TaskRegr$new(new_data, target = "biomass", id = "prediction")
set.seed(123) data_split = partition(task_selected, ratio = 0.8) train_indices = data_split$train test_indices = data_split$test
train_data = task_selected$data()[train_indices] train_task = TaskRegr$new( as.data.table(train_data), target = "biomass", id = "prediction_train" )
test_data <- task_selected$data()[test_indices] test_task = TaskRegr$new( as.data.table(test_data), target = "biomass", id = "prediction_test" )
learner_ranger = lrn("regr.ranger", id = "ranger")
learner_ranger$train(train_task) learner_ranger$model
resampling = rsmp("cv", folds=10) rr = resample(train_task, learner = learner_ranger, resampling = resampling) print(rr$aggregate())
predictions = learner_ranger$predict(test_task) measures = list(msr("regr.rmse"), msr("regr.mae"), msr("regr.rsq")) performance = predictions$score(measures) print(performance)
learner = lrn("regr.ranger", mtry = to_tune(lower = 1, upper = 9, logscale = FALSE), num.trees = to_tune(lower = 50, upper = 1500, logscale = FALSE), min.node.size = to_tune(lower = 1, upper = 20, logscale = FALSE) )
instance = ti( task = train_task, learner = learner, resampling = rsmp("cv", folds = 5), measures = msr("regr.mae"), terminator = trm("none") ) instance tuner = tnr("random_search", resolution = 5)
Error in assert_binary(truth, response = response, positive = positive, : Assertion on 'truth' failed: Must be of type 'factor', not 'character'.
The code is as above, and the error is reported when running to the tuner function. I am creating a regression task, it is reasonable that the data are numeric type, the whole process does not need factor type data, but the error says that it needs factor type data.