我有如下所示的结构化数据:
> str(df.train)'data.frame': 124 obs. of 11 variables: $ predictor01 : int 6 3 2 10 1 5 0 1 11 9 ... $ predictor02 : int 148 78 197 168 189 166 118 115 143 171 ... $ predictor03 : int 72 50 70 74 60 72 84 70 94 110 ... $ predictor04 : int 35 32 45 0 23 19 47 30 33 24 ... $ predictor05 : int 0 88 543 0 846 175 230 96 146 240 ... $ predictor06 : num 33.6 31 30.5 38 30.1 25.8 45.8 34.6 36.6 45.4 ... $ predictor07 : num 0.627 0.248 0.158 0.537 0.398 0.587 0.551 0.529 0.254 0.721 ... $ predictor08 : int 50 26 53 34 59 51 31 32 51 54 ... $ class-target : Factor w/ 2 levels "Sem_diabetes",..: 2 2 2 2 2 2 2 2 2 2 ... $ class-target-binaray1 : logi TRUE TRUE TRUE TRUE TRUE TRUE ... $ class-target-binaray2 : logi FALSE FALSE FALSE FALSE FALSE FALSE ...这是它们的摘要:
predictor01 predictor02 predictor03 predictor04 predictor05 predictor06 Min. : 0.000000 Min. : 0.0000 Min. : 0.00000 Min. : 0.00000 Min. : 0.00000 Min. : 0.00000 1st Qu.: 1.000000 1st Qu.:102.0000 1st Qu.: 64.75000 1st Qu.: 0.00000 1st Qu.: 0.00000 1st Qu.:26.97500 Median : 3.000000 Median :124.0000 Median : 72.00000 Median :25.00000 Median : 0.00000 Median :33.20000 Mean : 3.725806 Mean :127.6371 Mean : 69.98387 Mean :22.37903 Mean : 78.41129 Mean :32.90242 3rd Qu.: 6.000000 3rd Qu.:151.2500 3rd Qu.: 80.50000 3rd Qu.:34.00000 3rd Qu.:131.25000 3rd Qu.:37.40000 Max. :14.000000 Max. :197.0000 Max. :110.00000 Max. :63.00000 Max. :846.00000 Max. :67.10000 predictor07 predictor08 class-target class-target-binaray1 class-target-binaray2 Min. :0.078000 Min. :21.00000 Target1 :62 Mode :logical Mode :logical 1st Qu.:0.262500 1st Qu.:23.75000 Target2 :62 FALSE:62 FALSE:62 Median :0.407000 Median :27.50000 TRUE :62 TRUE :62 Mean :0.495871 Mean :32.20161 3rd Qu.:0.657500 3rd Qu.:39.50000 Max. :2.420000 Max. :67.00000 我通过以下方式创建了模板:
mod.rn <- neuralnet( class-target-binaray1 + class-target-binaray2 ~ predictor01 + predictor03 + predictor04 + predictor05 + predictor06 + predictor07, df.train, hidden=4, linear.output = TRUE, stepmax=1e6)到目前为止一切顺利。
然后，我将生成的模型应用于测试数据，执行如下操作:
test_result <- compute(mod.rn, df.test[,1:8])测试数据的结构与训练数据相同。所以我使用[,1:8]来仅选择预测变量。
当我运行上述最后一条命令时，收到了以下错误消息:
Error in neurons[[i]] %*% weights[[i]] : argumentos não compatíveis有什么想法吗?
回答：
看起来你没有用8个预测变量来拟合你的模型 – predictor02和predictor08在你的neuranet()中缺失了。
尝试明确地只保留相关预测变量，即:
myvars <- c('predictor01', 'predictor03', 'predictor04', 'predictor05', 'predictor06', 'predictor07')test_result <- compute(mod.rn, df.test[myvars])或者，如果你只是误漏了这两个预测变量，将它们加回你的neuralnet()定义中，你的代码应该可以正常工作。
