# Understanding Precision, Recall, F1-score and Confusion Matrix.

I believe it is a common topic. However, I am sharing how I understood these concepts because it took me a while to grasp these concepts.

First, I will start with a **confusion matrix** which is also known as an Error matrix. This is a fundamental concept to understand in the field of machine learning.

**Confusion matrix**displays the number of true positives, true negatives, false positives, and false negatives given some number of input data points (*usually*`n`

)

From the image above, `n`

is the *input data points*. Here we have 2 classes which are :

- Actual No: Class 1
- Actual Yes: Class 2

Confusion matrix helps you understand the

precisionandrecallof your model. How? I will come back to this. Let's talk aboutprecisionandrecall.

Precision and Recall are just different metrics for measuring the “success” or performance of a trained model.

**Precision** shows how often the models correct when it predicts the positive label. That is the number of true positives (in this case) over *all* positives, and will be the higher when the amount of false positives is low. The formula is **T**rue **P**ositives (**TP**) over Total Predicted Positives (**TP + FP**). **FP** means **F**alse **P**ositive.

False Positiveis where the model incorrectly indicates the presence of a condition.

Checking the example, the actual is **No** but the model predicted **Yes**.

So in the above example, that would be **45** over **60** (*where 45 is TP and FP is 15 = 60 *)for a **precision** score of 75 percent.

**Recall** shows how many of the **T**rue **P**ositives your model predicted. The formula is **T**rue **P**ositives over Total Real Positives(**TP** + **FN**). That is the number of true positives over true positives *plus* false negatives and will be higher when the number of false negatives is low.

False Negativemeans where the model incorrectly fails to indicate the presence of a condition when it is present. (read Type I & II error)

So here that’s **45** out of **50** for a score of 90 percent for a recall.

Going back to what I said before, how confusion matrix helps you understand the precision and recall of your model ?

Simply, it helps by counting out the True and False positives, as well as the True and False negatives.

Next let’s talk about F1-Score.

The **F1 score** is quite a combination of precision and recall. Because there are often trade-offs and aiming for higher precision or higher recall, you can aim to increase the F1 score.

The formula is shown in the image below.

F1 score at its highest is 1 indicating perfect precision and recall, at its lowest possible value is 0.

In the example given at the beginning of the article, the F1 score for the example is 0.83 approximately.

Thank you for reading.