The motivation of this post is to aide the understanding of the classification technique by explaining the evaluation tool for classification experiments: the confusion matrix. The name certainly does not inspire confidence in understanding, but let’s set that aside for a moment.
Lets consider a simple classification example: a machine learning experiment to predict if a customer will churn.
Recall that with supervised machine learning, a data set of examples (called the training set) is used for the machine to learn examples and explanatory relationships. The confusion matrix is a 2 x 2 grid with 4 measures that illustrate how effective the model performed on the test set of data. The four metrics are
- True Positive
- True Negative
- False Positive
- False Negative