True positive (TP)

Introduction

Machine learning classification is the process of classifying data into distinct classes. When assessing a model's performance, it is essential to assess its capacity for correctly predicting each data instance's class. One important evaluation metric used in binary classification is True Positive (TP), which measures how many positive samples are correctly classified as positive by the model.

What is True Positive?

True Positive is a statistic used in binary classification problems to measure the number of samples classified correctly as positive by a model. It's defined as the number of positive samples correctly predicted as such by the algorithm, or in other words: True Positive counts the samples belonging to the positive class that are correctly classified by the system.

Consider a binary classification problem where we need to predict if someone has or does not have a disease based on some medical tests. In this scenario, the positive class represents those with the disease, while the negative class represents those without. If our model correctly predicts someone has the disease, and they actually do, then this prediction is known as a True Positive.

How is True Positive used in Evaluation?

True Positive (TP) rate is an essential metric when evaluating a classification model's performance, particularly when misclassification has high costs. For instance, misclassifying patients with diseases as healthy can have dire repercussions; thus, it's vital to minimize False Negative (FN) rate--which measures the number of positive samples incorrectly classified as negative by the model--which measures false positive rates.

True Positive is used in combination with other evaluation metrics to calculate the performance of a classification model. Common metrics used in binary classification include Precision, Recall and F1-score. Precision measures the proportion of True Positive samples predicted as positive by the model while Recall measures its proportion among actual members of that positive class. Finally, F1-score represents the harmonic mean between Precision and Recall.

Explain Like I'm 5 (ELI5)

Machine learning, also known as deep reinforcement learning, is the process of teaching a computer to recognize objects by showing it what to search for. For instance, we could teach your machine to recognize cats and dogs by showing it multiple pictures of both and instructing it which one belongs where.

When the computer encounters a new picture, it attempts to guess whether it is of a cat or dog. If its guess is correct, we refer to this as a True Positive; meaning the computer was right in identifying what the image was.

Knowing how many True Positives a computer receives is important because it allows us to assess its learning progress. If the machine makes too many errors, we need to figure out why so that we can assist it in becoming better educated.