multiClassMetrics: Get's class-specific evaluation metrics for a multiclass...

In wtcooper/modUtils: Collection of utils for model building and tuning.

Description

Get's class-specific evaluation metrics for a multiclass classification model. For each class, includes sensitivity, specificity, and balanced accuracy. Also includes calculation of a custom balanced accuracy that is the product of a specific labels balanced accuracy (mean of spec and sens) times the minimum of spec or sens.

Designed to work as caret summaryFunction where data is a data.frame with columns 'pred' and 'obs' for predicted and observed labels.

Usage

multiClassMetrics(data, lev = NULL, model = NULL)

Arguments

data	data.frame with columns 'pred' and 'obs' for predicted and observed labels
lev	caret summaryFunction optional parameter
model	caret summaryFunction optional parameter

wtcooper/modUtils documentation built on May 4, 2019, 11:59 a.m.