Confusing metrics around the Confusion Matrix

"If you can't measure it, you can't possibly improve it" . In the field of Machine Learning and Data Science, especially with statistical classification, a "Confusion Matrix" is often used to derive a bunch of metrics that can be examined to either improve the performance of a classifier model or to compare the performance of multiple models. Instead of starting from the mathematical formulae for the metrics, we will try to intuitively derive the formulae based on basic concepts. It is probably called "confusion" because it depicts how much confused the classifier was while doing its predictions -- some classes were correctly classified and some were not. The most important concept to understand before exploring any metric from the confusion matrix is the true meaning of the "positive" and the "negative" class in the context of the problem given to the classifier. The Positive class is the existence what we are trying to detect or predict.