Question

我的数据（内部）具有3个类和3个自变量（类型，份额，值）的因变量（类）。我将其划分为训练和测试数据，然后将多类逻辑回归应用于训练数据集和对测试数据集的预测。我的准确率是34％。现在，我想使用“ multiclass.roc函数”计算曲线下的面积，但是我做不到。

library(pROC)

library(nnet)

summary(insider)

#class            type            shares              value          
 buyuk:200000   al :106805   Min.   :0.000e+00   Min.   :0.000e+00  
 kucuk:150000   sat:393195   1st Qu.:3.000e+02   1st Qu.:6.651e+03  
 orta :150000                Median :1.250e+03   Median :2.810e+04  
                             Mean   :2.963e+07   Mean   :5.071e+11  
                             3rd Qu.:6.800e+03   3rd Qu.:1.793e+05  
                             Max.   :5.774e+12   Max.   :9.677e+14


set.seed(7267166)
trainIndex=createDataPartition(insider$class, p=0.7)$Resample1
train=insider[trainIndex, ]
test=insider[-trainIndex, ]
model <- nnet::multinom(class ~., data = train)
predicted.classes <- model %>% predict(test)
mean(predicted.classes == test$class)

[1] 0.338

multiclass.roc(test$class, predict(model, test, type = 'prob'))

#Error in roc.default(response, predictor, levels = X, percent = percent, : Response and predictor must be vectors of the same length. In addition: Warning message: In roc.default(response, predictor, levels = X, percent = percent, : Deprecated use a matrix as predictor. Unexpected results may be produced, please pass a numeric vector.

在R中使用multiclass.roc函数的示例

0 个答案: