binaryClassifierEvaluation: Evaluate a Binary Classification Model
In ivanliu1989/RQuant: A collection of generic functions

Description Usage Arguments Value See Also Examples

Evaluate the performance of a model with two outcomes, 0 and 1.

Given a vector of estimated probabilities of success and a vector of actual successes and failures, this function creates some standardised graphs and summaries and returns them as a list that can be consumed by other functions such as binaryClassifierEmailReport or a shiny app.

The result contains a histogram of estimated probabilities, the ROC plot with the calculated area under the curve, two density plots, one violin, one standard, two loss curves as a function of deciles of the calculated probabilies, and a plot of cumulative It contains some stats such as Area under the ROC curve, Root Mean Square Error and the confusion matrix. It does not require the actual model. If the model is supplied, a summary of the model is also included in the output.

1 2	binaryClassifierEvaluation(probs, actuals, thresholdConfusion = median(probs), model = NULL, config = list(finish = Sys.time(), n.test = length(probs)))

`probs`	A vector of estimated probabilities, from 0 to 1.
`actuals`	A vector of actual outcomes, 0 or 1.
`thresholdConfusion`	a threshold to convert probabilities to 0's or 1's. Takes the median of the `probs` as a default.
`model`	Optional, the model used to estimate `probs` from `actuals`
`config`	A list containing details of the call to the function, such as lenght of vectors or time of completion. This is simply printed and attached to the email.

A list with an additional class binaryModelEvaluation

binaryClassifierEmailReport

set.seed(123)
probs <- pnorm(rnorm(100,-1,1))
hist(probs)
actuals <- rbinom(100,1,probs)
plot(as.factor(actuals),probs)

binaryClassifierEvaluation(probs, actuals)

my.data<-data.frame(actuals,probs)
mod.fit <- glm(actuals ~ probs , data=my.data, family=binomial)

binaryClassifierEvaluation(probs, actuals
   , model = mod.fit
   , config = list(
       finish = Sys.time()
       ,n.text = length(probs)
       )
   )