binary_model_evaluation: Calculate Binary Model Performance Metrics

In mattmills49/modeler: Exploratory Data Analysis and Modeling Helper Functions

Description

Accepts a data frame containing the model predictions and actual values. The following classification variables are calculated on the predictions:

Usage

binary_model_evaluation(.data, prediction_formula, group_var = NULL)

Arguments

.data	a data frame
prediction_formula	a formula specifiying the dependent binary variable (lhs) and the probability predictions (rhs)
group_var	an unquoted variable name to calculate performance measures on each subset of the data - e.g., training/testing splits

Details

• AUC - Area Under the Receiver Operating Curve

• KS - Kolmogorov-Smirnov statistic

• MSE - Mean Square Error

• TPR - True Positive Rate

• TNR - True Negative Rate

• LSR - Logistic Scoring Rule (Log Loss)

• Bias - Direction of Bias

Value

a data frame containing the metrics

Examples

iris_ <- iris
iris_$setosa <- ifelse(iris_$Species == "setosa", 1, 0)
iris_$pred <- iris_$Sepal.Length
iris_$pred <- (max(iris_$pred) - iris_$pred) / (max(iris_$pred) - min(iris_$pred) + 1)
iris_$splits <- sample(c("Test", "Train"), size = nrow(iris_), replace = T)
binary_model_evaluation(iris_, setosa ~ pred)
binary_model_evaluation(iris_, setosa ~ pred, group_var = splits)

mattmills49/modeler documentation built on May 21, 2019, 1:25 p.m.