In this article we explain you **THE method to never forget** the difference between True Positives, False Negatives, True Negatives and False Positives.

These terms define the type of result we can get after **a prediction with two possibilities** (in Machine Learning we call it binary classification)

They enable to** evaluate the performance of our model** and know how reliable its predictions are.

Often thes**e terms are confused** with each other and that’s why we give you here **the technique to remember them !**

## Why use True Positives / False Negatives ?

**To measure the performance of a Machine Learning model**, we cannot simply look at the number of well-made predictions.

On the contrary, this approach can give us **a distorted view of our model**.

Let’s take the example of **a choice with two possibilities** : we place ourselves **at the exit of a tunnel to predict what will come out.** We imagine that there are two possibilities here, either it is **a car** either **a motorcycle**.

If on this road **80% of the users are in cars** and **we predict each time that it will be a car** that will come out of the tunnel… We will have at the end **a success rate of 80%.**

This is what we call **a naive prediction** because there is no real reflection: we don’t predict, we arbitrarily decide that everything that comes out of the tunnel will be a car.

The problem is that this approach works when there are 80% of cars on the road, but **if the context changes, this approach will not work anymore.**

In fact, this type of prediction is not reliable. Indeed, **since our data is biased** (there are many more cars than bikes), **our performance is biased too.**

That’s why **a naive approach is not efficient**. We need **an smart approach.**

What we are trying to do in **Machine Learning** is to have an smart approach, **a model that can be reproduced in different contexts** (in our case, on different tunnels).

Fortunately for us, there is **a performance analysis** that allows us to give **more importance to smart models** than to naive models.

## How True Positives / False Negatives work?

This method is called **True Positives/False Negatives.**

Let’s go back to our **tunnel example.**

We have here **two choices :**

**a car**comes out of the tunnel**a motorcycle**comes out of the tunnel

The objective here is **to measure the performance** of our prediction on these two choices.

By default, we’ll say that the **car** corresponds to the “**positive**” choice and the **motorcycle** to the “**negative**” choice.

**For each of these choices**, either our prediction is **true** (good) or it is **false** (bad).

For example, **we predict that a car will leave the tunnel**. We place ourselves in the “**positive**” choice.

**A car does come out of the tunnel**, so our prediction is **true**. It is a **True Positive**.

If, on the other hand, our prediction was **wrong** (a motorcycle comes out of the tunnel), we say that it is a **False Positive.**

Same thing for the “**negative**” choice: **we predict that it is a motorcycle that will come out of the tunnel.**

**A motorcycle comes out**, so our prediction is **true**. It is a **True Negative**.

But** if a car comes out**, our prediction is **false**. It is a **False Negative**.

Finally, we have **four possible results :**

- a car comes out of the tunnel
- we predicted that a car would come out –
**True Positive** - we predicted a motorcycle would come out –
**False Negative**

- we predicted that a car would come out –
- a motorcycle comes out of the tunnel
- we predicted a car would come out –
**False Positive** - we predicted a motorcycle would come out –
**True Negative**

- we predicted a car would come out –

Once seen, it is easy **to conflate these concepts.** That’s why we offer you **two methods to remember them easily !**

## Written method

Here we must keep in mind that :

**car = positive****motorcycle = negative**

Then, for each prediction, **we use the sentence** “*We predicted …(2) and it was …(1)*“.

For example, if **we predicted that a car would come out of the tunnel**, but in the end it was a motorcycle that came out, we would have :

“*We predicted positive (2) and it was false (1)*“.

We take **(1)** and** (2)** which gives us : **False Positive.**

Said in **a more detailed way:**

“We predicted that a **car (2)** would come out of the tunnel and it was ultimately **false (1)**“.

We take **(1)** and **(2)** which gives us **False Car** but as said at the beginning of the exercise car = positive so, we have **False Positive**.

The method is therefore to remember this sentence “We predicted **…(2)** and it was **…(1)**” and to complete it according to the prediction.

## Visual method

Here again, keep in mind the basic assumption is :

**car = positive****motorcycle = negative**

For example : **we predicted that a motorcycle will come out of the tunnel** and finally **it is indeed a motorcycle that comes out.**

In this case our prediction is **true (1)** and we predicted that it was a **motorcycle = negative (2)**.

We take **(1)** and **(2)** which gives us : **True Negative.**

For this method you only need to **keep this table in mind** and **fill in the middle column with the corresponding choices** (in our case “car” and “motorcycle”).

## To Know

In **Machine Learning**, in **order to display the result** of a prediction model with two possibilities (like the example we just saw), we use **a confusion matrix**.

The confusion matrix is in fact **a table in which we display the number of predictions** according to each possibility:

True Positives number | False Negatives number |

False Positives number | True Negatives number |

**You may then ask me **: is the confusion matrix actually just a fancy word for “result in table form”?

I would answer that yes, it is. There are people who like to make things complicated for no reason, but it still is **a used term in Machine Learning !**

Remember that the confusion matrix is **a table that displays the results of the model.**

**To go further** and if you want to know more about **how to effectively use True Positives and False Negatives** you can continue the reading with this post on Recall, Accuracy and F1 Score ! 🙂

**sources :**

- L. Antiga,
*Deep Learning with PyTorch*(2020, Manning Publications) – our affiliated link - Photo by Ryan Millsap on Unsplash