StatMeasures (version 1.0)

accuracy: Confusion matrix and overall accuracy of predicted binary response

Description

Takes in actual binary response, predicted probabilities and cutoff value, and returns confusion matrix and overall accuracy

Usage

accuracy(y, yhat, cutoff)

Arguments

y

actual binary response variable

yhat

predicted probabilities corresponding to the actual binary response

cutoff

threshold value in the range 0 to 1

Value

a three element list: confusion matrix as a table, confusion matrix (percentages) as a table and overall accuracy value

Details

When we predict a binary response, first thing that we want to check is accuracy of the model for a particular cutoff value. This function does just that and provides confusion matrix (numbers and percentage) and overall accuracy. Overall accuracy is calculated as (TP + TN)/(P + N).

The output is a list from which the individual elements can be picked as shown in the example.

See Also

ks, auc, iv, splitdata

Examples

# A 'data.frame' with y and yhat
df <- data.frame(y = c(1, 0, 1, 1, 0),
                 yhat = c(0.86, 0.23, 0.65, 0.92, 0.37))

# Accuracy tables and overall accuracy figures
ltAccuracy <- accuracy(y = df[, 'y'], yhat = df[, 'yhat'], cutoff = 0.7)
accuracyNumber <- ltAccuracy$accuracyNum
accuracyPercentage <- ltAccuracy$accuracyPer
overallAccuracy <- ltAccuracy$overallAcc