当 Resample 的AUC结果与基准(bench)在 均使用Bootstrap方法 的场景下出现不一致时>…
一个查到的方案是:在学习器添加随机种子
learner$param_set$values$seed = 123
Benchmark
set.seed(123)
design = benchmark_grid(task,learners ,
rsmps("bootstrap", repeats =200))
future::plan("multisession",workers =7)
bmr = benchmark(design, store_models = T,store_backends = T)
autoplot(bmr, type = "roc")
Resample
set.seed(123)
rr <- resample(task, learner_rf,
resampling = rsmp("bootstrap",repeats=200),
store_models = T,store_backends = T)
rr$aggregate(msr("classif.auc"))