accuracy_threshold: Accuracy Threshold

In ChrisDienes/SeaClass: Interactive GUI for analyzing a binary response variable

Description

Returns the best binary threshold rule based on accuracy.

Usage

accuracy_threshold(x, group, pos_class)

Arguments

x	a numeric predictor.
group	binary grouping variable.
pos_class	group level for positive class.

Details

Optimizes threshold rule based on misclassification rate (1 - accuracy). Upper rules are defined as greater than cut, and lower is less than or equal to cut.

Value

• misclass rate of misclassification.

• cut selected threshold.

• direction indicator for positive group direction.

See Also

max_fp_threshold, score_threshold

Examples

### General Use: ###
set.seed(123)
x <- c(rnorm(100,0,1),rnorm(100,2,1))
group <- c(rep(0,100),rep(2,100))
accuracy_threshold(x=x, group=group, pos_class=2)
accuracy_threshold(x=x, group=group, pos_class=0)
### Bagged Example ###
set.seed(123)
replicate_function = function(index){accuracy_threshold(x=x[index], group=group[index], pos_class=2)$cut}
sample_cuts = replicate(100, {
  sample_index = sample.int(n=length(x),replace=TRUE)
  replicate_function(index=sample_index)
})
bagged_scores = sapply(x, function(x) mean(x > sample_cuts))
unbagged_cut    = accuracy_threshold(x=x, group=group, pos_class=2)$cut
unbagged_scores = ifelse(x > unbagged_cut, 1, 0)
# Compare AUC:
PRROC::roc.curve(scores.class0 = bagged_scores,weights.class0 = ifelse(group==2,1,0))$auc
PRROC::roc.curve(scores.class0 = unbagged_scores,weights.class0 = ifelse(group==2,1,0))$auc
bagged_prediction = ifelse(bagged_scores > 0.50, 2, 0)
unbagged_prediction = ifelse(x > unbagged_cut, 2, 0)
# Compare Confusion Matrix:
table(bagged_prediction, group)
table(unbagged_prediction, group)

ChrisDienes/SeaClass documentation built on May 18, 2019, 9:15 p.m.