statistics: Calculation of different performance measures and statistical...

In UBESP-DCTV/costumer: COmprehensive Searches ThroUgh Machine learning for systEmatic Reviews

Calculation of different performance measures and statistical values

Description

statistics returns the different performance measures and statical values using the function confusionMatrix and the package pROC for the AUC-ROC calculation

Usage

statistics(pred, truth, positive, dnn, prevalence, mode)

Arguments

pred	factor containing the predicted classes
truth	factor containing the reference classes
positive	the first level of the factor or the level considered as positive class, if factors are classified as positive and negative classes
dnn	dimension names for the table
prevalence	prevalence should be a single numeric value since we are using the binary factors
mode	specifies either particular statistical values or everything returns all statistical values

Value

list containing two elements. First element is a list with elements: table (confusion matrix), positive (the level of positive class), overall (overall accuracy and other statistic values) and byClass (the values of different performance measures as specified by the argument mode). Second element is the value: Area under the curve

Examples

library(tm)
library(plyr)

truth <- factor(meta(liu_corpus)$real_label)
sam   <- meta(liu_corpus)$real_label
pred  <- factor(sample(sam))
statistics(pred, truth, positive = "1",  dnn = c("Prediction", "Truth"),
  prevalence = NULL, mode = "everything"
)

UBESP-DCTV/costumer documentation built on Feb. 1, 2023, 4:52 a.m.