R language - some metrics for unbalanced data sets

R Language - Some metrics for unbalanced data sets

In view of the fact that many posts have introduced various indicators to measure the performance of classifiers in detail , Here I want to share with you the functions written by the author , Function will return F_score（F Measure , Default β take 1）,BACC（balanced accuracy）,recall Recall rate , precision Accuracy ,specifity What a negative rate

The code is as follows ：

# Calculate some indicators of unbalanced data sets , The main output F-score and BACC
# here positiveclass and  negativeclass Value is the value of the target attribute , Such as yes and no
UBMetrics<-function(preds, test.lable, positiveclass, negativeclass, bta=1){
  recall<-length(which(preds==test.label&preds==positiveclass))/length(which(test.label==positiveclass))
  
  precision<-length(which(preds==test.label&preds==positiveclass))/length(which(preds==positiveclass))
  
  specifity<-length(which(preds==test.label&preds==positiveclass))/length(which(test.label==positiveclass))
  
  F_score<-(1+bta^2)*precision*recall/(recall+precision)
  BACC<-(recall+specifity)/2
  
  return(c(F_score=F_score,BACC=BACC,recall=recall,precision=precision,specifity=specifity))
}

The above code generation script can be named arbitrarily , It's called UBMetrics, Reference when calling ：
source("D:\\workspace\\UBMetrics.R")