Classification

Logistic Regression

Machine Learning

January 7, 2021

Explain the significance of ROC.

Receiver Operator Curve (ROC) is used for finding the optimum threshold for Sensitivity. Sensitivity or Recall is a measure of rate of ‘True Positives’. It plots Sensitivity against rate of ‘False Positives’ i.e. against (1-Specificity). Research has shown that up till a certain threshold both ‘True Positive’ rate and ‘False Positive’ rate increases but beyond that ‘True Positive’ rate becomes constant i.e. indifferent to increments in ‘False Positive’ rate.

Thus identifying this threshold becomes imperative for selecting/identifying the best hyperparameters for a given algorithm. In some business problems you want to minimize ‘False Positives’, even at the expense capability to identify ‘True Positives’. In such cases ROC curve helps you to stem the ‘False Negative’ rate at the desired value.

by : Monis Khan

Quick Summary:

Receiver Operator Curve (ROC) is used for finding the optimum threshold for Sensitivity. Sensitivity or Recall is a measure of rate of ‘True Positives’. It plots Sensitivity against rate of ‘False Positives’ i.e. against (1-Specificity). Research has shown that up till a certain threshold both ‘True Positive’ rate and ‘False Positive’ rate increases but beyond […]