在R中查找精确度、召回率和F1分数。

huangapple
go评论93阅读模式
英文:

Finding the precision, recall and the f1 in R

问题

I want to run models on a loop via and then store the performance metrics into a table. I do not want to use the confusionMatrix function in caret, but I want to compute the precision, recall and f1 and then store those in a table. Please assist, edits to the code are welcome.
My attempt is below.

  1. library(MASS) #will load our biopsy data
  2. library(caret)
  3. data("biopsy")
  4. biopsy$ID <- NULL
  5. names(biopsy) <- c('clump thickness','uniformity cell size','uniformity cell shape',
  6.                  'marginal adhesion','single epithelial cell size','bare nuclei',
  7.                  'bland chromatin','normal nuclei','mitosis','class')
  9. sum(is.na(biopsy))
  10. biopsy <- na.omit(biopsy)
  11. sum(is.na(biopsy))
  12. head(biopsy, 5)
  13. set.seed(123)
  14. inTraining <- createDataPartition(biopsy$class, p = .75, list = FALSE)
  15. training <- biopsy[inTraining,]
  16. testing  <- biopsy[-inTraining,]
  18. # Run algorithms using 10-fold cross validation
  19. control <- trainControl(method="repeatedcv", number=10, repeats = 5, verboseIter = FALSE, classProbs = TRUE)
  20. # CHANGING THE CHARACTERS INTO FACTORS VARIABLES
  21. training <- as.data.frame(unclass(training),                     
  22.                          stringsAsFactors = TRUE)
  24. # CHANGING THE CHARACTERS INTO FACTORS VARIABLES
  25. testing <- as.data.frame(unclass(testing),                     
  26.                          stringsAsFactors = TRUE)
  28. models <- c("svmRadial", "rf")
  29. results_table <- data.frame(models = models, stringsAsFactors = FALSE)
  31. for (i in models){
  32.   model_train <- train(class ~ ., data = training, method = i,
  33.                      trControl = control, metric = "Accuracy")
  34.   predictions <- predict(model_train, newdata = testing)
  35.   precision_ <- posPredValue(predictions, testing)
  36.   recall_ <- sensitivity(predictions, testing)
  37.   f1 <- (2 * precision_ * recall_) / (precision_ + recall_)
  39.   # put that in the results table
  40.   results_table[i, "Precision"] <- precision_
  41.   results_table[i, "Recall"] <- recall_
  42.   results_table[i, "F1score"] <- f1
  43. }

However, I get an error which says Error in posPredValue.default(predictions, testing) : inputs must be factors. I do not know where I went wrong, and any edits to my code are welcome.

I know that I could get precision, recall, f1 by just using the code below (B), however, this is a tutorial question where I am required not to use the code example below (B):

  1. (B)
  3. for (i in models){
  4.   model_train <- train(class ~ ., data = training, method = i,
  5.                      trControl = control, metric = "Accuracy")
  6.   predictions <- predict(model_train, newdata = testing)
  7.   print(confusionMatrix(predictions, testing$class, mode = "prec_recall"))
  8. }
英文:

I want to run models on a loop via and then store the performance metrics into a table. I do not want to use the confusionMatrix function in caret, but I want to compute the precision, recall and f1 and then store those in a table. Please assist, edits to the code are welcome.
My attempt is below.

  1. library(MASS) #will load our biopsy data
  2. library(caret)
  3. data(&quot;biopsy&quot;)
  4. biopsy$ID&lt;-NULL
  5. names(biopsy)&lt;-c(&#39;clump thickness&#39;,&#39;uniformity cell size&#39;,&#39;uniformity cell shape&#39;,
  6.                  &#39;marginal adhesion&#39;,&#39;single epithelial cell size&#39;,&#39;bare nuclei&#39;,
  7.                  &#39;bland chromatin&#39;,&#39;normal nuclei&#39;,&#39;mitosis&#39;,&#39;class&#39;)
  9. sum(is.na(biopsy))
  10. biopsy&lt;-na.omit(biopsy)
  11. sum(is.na(biopsy))
  12. head(biopsy,5)
  13. set.seed(123)
  14. inTraining &lt;- createDataPartition(biopsy$class, p = .75, list = FALSE)
  15. training &lt;- biopsy[ inTraining,]
  16. testing  &lt;- biopsy[-inTraining,]
  18. # Run algorithms using 10-fold cross validation
  19. control &lt;- trainControl(method=&quot;repeatedcv&quot;, number=10,repeats = 5, verboseIter = F, classProbs = T)
  20. #CHANGING THE CHARACTERS INTO FACTORS VARAIBLES
  21. training&lt;- as.data.frame(unclass(training),                     
  22.                          stringsAsFactors = TRUE)
  24. #CHANGING THE CHARACTERS INTO FACTORS VARAIBLES
  25. testing &lt;- as.data.frame(unclass(testing),                     
  26.                          stringsAsFactors = TRUE)
  29. models&lt;-c(&quot;svmRadial&quot;,&quot;rf&quot;)
  30. results_table &lt;- data.frame(models = models, stringsAsFactors = F)
  32. for (i in models){
  33.   model_train&lt;-train(class~., data=training, method=i,
  34.                      trControl=control,metric=&quot;Accuracy&quot;)
  35.   predictions&lt;-predict(model_train, newdata=testing)
  36.   precision_&lt;-posPredValue(predictions,testing)
  37.   recall_&lt;-sensitivity(predictions,testing)
  38.   f1&lt;-(2*precision_*recall_)/(precision_+recall_)
  40.   # put that in the results table
  41.   results_table[i, &quot;Precision&quot;] &lt;- precision_
  42.   results_table[i, &quot;Recall&quot;] &lt;- recall_
  43.   results_table[i, &quot;F1score&quot;] &lt;- f1
  44. }

However I get an error which says Error in posPredValue.default(predictions, testing) :
inputs must be factors. i do not know where I went wrong and any edits to my code are welcome.

I know that I could get precision,recall, f1 by just using the code below (B), however this is a tutorial question where I am required not to use the code example below (B):

  1. (B)
  3. for (i in models){
  4.   model_train&lt;-train(class~., data=training, method=i,
  5.                      trControl=control,metric=&quot;Accuracy&quot;)
  6.   predictions&lt;-predict(model_train, newdata=testing)
  7.   print(confusionMatrix(predictions, testing$class,mode=&quot;prec_recall&quot;))
  8. }

答案1

得分: 1

需要发生一些事情。

  1. 您需要更改posPredValue和sensitivity的函数调用。对于两者，将testing更改为testing$class。

  2. 对于results_table，i是一个_单词_，而不是一个值，所以您正在分配results_table["rf", "Precision"] <- precision_（这会创建一个新行，行名为"rf"）。

以下是您的for语句，其中包括对1)中提到的函数的更改以及解决2)中问题的修改。

  1. for (i in models){
  2.   model_train <- train(class~., data = training, method = i,
  3.                      trControl= control, metric = "Accuracy")
  4.   assign("fit", model_train)
  5.   predictions <- predict(model_train, newdata = testing)
  6.   precision_ <- posPredValue(predictions, testing$class)
  7.   recall_ <- sensitivity(predictions, testing$class)
  8.   f1 <- (2*precision_ * recall_) / (precision_ + recall_)
  9.   
  10.   # 将这些值放入结果表
  11.   results_table[results_table$models %in% i, "Precision"] <- precision_
  12.   results_table[results_table$models %in% i, "Recall"] <- recall_
  13.   results_table[results_table$models %in% i, "F1score"] <- f1
  14. }

这是对我而言的样子。

  1. results_table
  2. #      models Precision    Recall   F1score
  3. # 1 svmRadial 0.9722222 0.9459459 0.9589041
  4. # 2        rf 0.9732143 0.9819820 0.9775785
英文:

A few things need to happen.

  1. You have to change the function calls for posPredValue and sensitivity. For both, change testing to testing$class.

  2. for the results_table, i is a word, not a value, so you're assigning results_table[&quot;rf&quot;, &quot;Precision&quot;] &lt;- precision_ (This makes a new row, where the row name is "rf".)

Here is your for statement, with changes to those functions mentioned in 1) and a modification to address the issue in 2).

  1. for (i in models){
  2.   model_train &lt;- train(class~., data = training, method = i,
  3.                      trControl= control, metric = &quot;Accuracy&quot;)
  4.   assign(&quot;fit&quot;, model_train)
  5.   predictions &lt;- predict(model_train, newdata = testing)
  6.   precision_ &lt;-posPredValue(predictions, testing$class)
  7.   recall_ &lt;- sensitivity(predictions, testing$class)
  8.   f1 &lt;- (2*precision_ * recall_) / (precision_ + recall_)
  9.   
  10.   # put that in the results table
  11.   results_table[results_table$models %in% i, &quot;Precision&quot;] &lt;- precision_
  12.   results_table[results_table$models %in% i, &quot;Recall&quot;] &lt;- recall_
  13.   results_table[results_table$models %in% i, &quot;F1score&quot;] &lt;- f1
  14.   }

This is what it looks like for me.

  1. results_table
  2. #      models Precision    Recall   F1score
  3. # 1 svmRadial 0.9722222 0.9459459 0.9589041
  4. # 2        rf 0.9732143 0.9819820 0.9775785

huangapple
  • 本文由 发表于 2023年2月16日 19:36:33
  • 转载请务必保留本文链接：https://go.coder-hub.com/75471701.html
  • caret
  • classification
  • dataframe
  • machine-learning
  • 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