随机林性能不好（使用tunerf调谐）和训练用插入符号

我正在与我的随机森林表演作斗争。在测试数据上，它给了我0 Kappa和零灵敏度。我优化了几个参数，但再次遇到相同的问题：这就是我在以下代码中所做的：

# set up cross validation

cross.val <- trainControl(method="repeatedcv", number = 10, repeats = 10)

# searches for best mtry given training data 
bestmtry[[i]] <- tuneRF(rf_train[[i]][,c(-80)], rf_train_fac[[i]], stepfactor = 0.5, improve =      1e-5, ntree=200, trace = TRUE, plot = TRUE)
     mtry_list[[i]] = bestmtry[[i]][,1]
     mtry <-  bestmtry[[i]][,1]
     tuning[[i]] <- expand.grid( mtry = c(bestmtry[[i]][,1]) )

# train models
set.seed(200)
pred.mod[[i]] = train(rf_train[[i]][,c(-80)], rf_train_fac[[i]],
                      method = 'rf', 
                      weights = NULL,
                      metric = "Kappa",
                      trControl = cross.val,
                      tuneGrid = tuning[[i]], 
                      importance = TRUE )

## make rf_test predictions ##   
rf_test_pred[[i]] <- predict(pred.mod[[i]], rf_test[[i]][,c(-80)]) 

# check accuracy and Kappa for rf_test predictions

CM_rf_test[[i]] <- confusionMatrix(rf_test_pred[[i]], rf_test_fac[[i]], positive = "1")

#设置交叉验证
cross.val嗨Jana，你能编辑你的问题来修正格式吗？底部的问题需要从行首删除空格。它触发了我的强迫症，不允许你编辑它