Measure Classifier Performance

The most starting point in classification performance is confusion matrix (or classification table). In confusion matrix consists the distribution of predicted value and actual value. It is tabled in the presentation of 2 by 2 matrix as shown below:

By using the confusion matrix (classification table), we can calculate a range of measures such as Accuracy, Precision, Sensitivity, Specificity, F-Score, and many more.

1) Accuracy is the overall proportion of correct classifications.

2) Precision is the proportion of predicted that were correct.

3) Sensitivity is the proportion of actual positives that were predicted correctly (sometimes called as recalled).

4) Specificity is the proportion of actual fails that were predicted correctly.

5) F-Score / F1-Score provides a balanced measure of precision and sensitivity.