classification_metrics: Classification metrics for assessing performance of...

In duncankmckinnon/OKdplyr: Zero Kelvin dplyr

Description

extract all performance metrics for a binomial classification model

Usage

classification_metrics(mod, y, y.hat, pi_0 = 0.5)

Arguments

mod	a general linear model of family binomial
y	actual classification values ( only used if mod not specified )
y.hat	model predicitions as percentage ( only used if mod not specified )
pi_0	the threshold of determination used in classifaction ( optional, default = 0.5 )

Value

A list of classification performance metrics derived from the confusion matrix

• y.hat : the model classification predictions

• confmatrix : the confusion matrix

• tpos : count of true positives

• tneg : count of true negatives

• fneg : count of false negatives

• fpos : count of false positives

• acc : accuracy score

• recall : recall score

• precision : precision score

• tpr : true positive rate (same as recall)

• fpr : false positive rate

• fdcr : false discovery rate

• fnr : false negative rate

• fomr : false omission rate

• tnr : true negative rate

• ppv : positive predicted value (same as precision)

• npv : negative predicted value

• F1 : F1 scoring metric

Examples

X <- rnorm(100)
Y <- ifelse(X < 0.5, 0, 1 - sample(0:1, prob = c(0.9, .1)))
f <- glm(Y ~ X, family='binomial')
classification_metrics(f)

duncankmckinnon/OKdplyr documentation built on Nov. 30, 2020, 1:13 a.m.