“Neuralnet”库在R中 – 混淆矩阵

huangapple
go评论88阅读模式
英文:

"Neuralnet" Library in R - Confusion Matrix

问题

我运行此代码时收到错误消息“所有参数必须具有相同的长度”。

是否有人可以向我展示如何克服这个错误，因为'test$y'和'pred_label'都具有相同的输入长度？

我目前正在尝试从这些数据中生成混淆矩阵（具有讽刺意味），并且将不胜感激地接受所有帮助。

  1. library(neuralnet)
  2. library(tidyverse)
  4. #MOCK DATA
  6. x1 = rep(1:3, times = 40)
  7. x2 = rep(1:3, times = 40)
  8. x3 = rep(1:3, times = 40)
  9. x4 = rep(1:3, times = 40)
  10. x5 = rep(1:3, times = 40)
  12. y = rep(0:1, times = 60)
  13. y <- as.factor(y)
  15. dat <- data.frame(y, x1, x2, x3, x4, x5)
  17. #SPLIT
  19. set.seed(123)
  21. indexes = createDataPartition(dat$y, p = .85, list = F)
  22. train = dat[indexes, ]
  23. test = dat[-indexes, ]
  25. xtest = test[, -1]
  26. ytest = test[, 1]
  28. #MODEL
  30. NN1 <- neuralnet(y ~., train,
  31.                  linear.output = FALSE,
  32.                  stepmax = 1e7)
  34. #ACCURACY TEST
  36. pred <- predict(NN1, test)
  37. labels <- c("0", "1")
  38. pred_label <- data.frame(max.col(pred)) %>%
  39.   mutate(pred = labels[max.col.pred.]) %>%
  40.   select(2) %>%
  41.   unlist()
  43. table(test$y, pred_label)

我目前正在遵循此教程：https://www.datacamp.com/tutorial/neural-network-models-r

英文:

I get an error message "all arguments must have the same length" when I run this code.

Would it be possible for someone to show me how to overcome this error, given `test$y,' and 'pred_label' both have the same input length?

I am currently struggling to generate a confusion matrix (ironically) from this data, and all help would be appreciated.

  1. library(neuralnet)
  2. library(tidyverse)
  5. #MOCK DATA
  7. x1 = rep(1:3, times = 40)
  8. x2 = rep(1:3, times = 40)
  9. x3 = rep(1:3, times = 40)
  10. x4 = rep(1:3, times = 40)
  11. x5 = rep(1:3, times = 40)
  13. y = rep(0:1, times = 60)
  14. y &lt;- as.factor(y)
  16. dat &lt;- data.frame(y, x1, x2, x3, x4, x5)
  18. #SPLIT
  20. set.seed(123)
  22. indexes=createDataPartition(dat$y, p=.85, list = F)
  23. train = dat[indexes, ]
  24. test = dat[-indexes, ] 
  26. xtest = test[, -1]
  27. ytest = test[, 1]
  29. #MODEL
  31. NN1 &lt;- neuralnet(y ~., train,
  32.                  linear.output = FALSE,
  33.                  stepmax=1e7)
  36. #ACCURACY TEST
  38. pred &lt;- predict(NN1, test)
  39. labels &lt;- c(&quot;0&quot;, &quot;1&quot;)
  40. pred_label &lt;- data.frame(max.col(pred)) %&gt;%     
  41.   mutate(pred=labels[max.col.pred.]) %&gt;%
  42.   select(2) %&gt;%
  43.   unlist()
  45. table(test$y, prediction_label)

I am currently following this tutorial. https://www.datacamp.com/tutorial/neural-network-models-r

答案1

得分: 1

你可以使用以下代码进行ROC计算

  1. library(MLeval)
  3. pred <- predict(NN1, newdata = test)
  4. colnames(pred) <- c("No", "Yes")
  6. roc <- data.frame(pred, y = test$y, Group = "ANN") %>%
  7.   mutate(y = if_else(y == 0, "No", "Yes"))
  8.   
  9. ROC <- evalm(roc)
英文:

You can use the following code

  1. library(neuralnet)
  2. library(tidyverse)
  3. library(caret)
  5. #MOCK DATA
  6. x1 = rep(1:3, times = 40)
  7. x2 = rep(1:3, times = 40)
  8. x3 = rep(1:3, times = 40)
  9. x4 = rep(1:3, times = 40)
  10. x5 = rep(1:3, times = 40)
  12. y = rep(0:1, times = 60)
  13. y &lt;- as.factor(y)
  15. dat &lt;- data.frame(y, x1, x2, x3, x4, x5)
  17. #SPLIT
  18. set.seed(123)
  19. indexes=createDataPartition(dat$y, p=.85, list = F)
  20. train = dat[indexes, ]
  21. test = dat[-indexes, ] 
  23. #MODEL
  24. NN1 &lt;- neuralnet(y ~., train,
  25.                  linear.output = FALSE,
  26.                  stepmax=1e7)
  29. #ACCURACY TEST
  30. pred &lt;- predict(NN1, newdata = test)
  31. colnames(pred) &lt;- c(&quot;0&quot;, &quot;1&quot;)
  32. tab &lt;- table(test$y, colnames(pred)[max.col(pred)])
  33. confusionMatrix(tab)
  35. #&gt; Confusion Matrix and Statistics
  36. #&gt; 
  37. #&gt;    
  38. #&gt;     0 1
  39. #&gt;   0 3 6
  40. #&gt;   1 6 3
  41. #&gt;                                           
  42. #&gt;                Accuracy : 0.3333          
  43. #&gt;                  95% CI : (0.1334, 0.5901)
  44. #&gt;     No Information Rate : 0.5             
  45. #&gt;     P-Value [Acc &gt; NIR] : 0.9519          
  46. #&gt;                                           
  47. #&gt;                   Kappa : -0.3333         
  48. #&gt;                                           
  49. #&gt;  Mcnemar&#39;s Test P-Value : 1.0000          
  50. #&gt;                                           
  51. #&gt;             Sensitivity : 0.3333          
  52. #&gt;             Specificity : 0.3333          
  53. #&gt;          Pos Pred Value : 0.3333          
  54. #&gt;          Neg Pred Value : 0.3333          
  55. #&gt;              Prevalence : 0.5000          
  56. #&gt;          Detection Rate : 0.1667          
  57. #&gt;    Detection Prevalence : 0.5000          
  58. #&gt;       Balanced Accuracy : 0.3333          
  59. #&gt;                                           
  60. #&gt;        &#39;Positive&#39; Class : 0

For ROC calculation, you can use the following code

  1. library(MLeval)
  3. pred &lt;- predict(NN1, newdata = test)
  4. colnames(pred) &lt;- c(&quot;No&quot;, &quot;Yes&quot;)
  6. roc &lt;- data.frame(pred, y = test$y, Group = &quot;ANN&quot;) %&gt;% 
  7.   mutate(y = if_else(y == 0, &quot;No&quot;, &quot;Yes&quot;))
  8.   
  9. ROC &lt;- evalm(roc)

huangapple
  • 本文由 发表于 2023年7月10日 20:18:13
  • 转载请务必保留本文链接：https://go.coder-hub.com/76653670.html
  • deep-learning
  • neural-network
  • r
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定