Classification, Regression, and Clustering
Compare three common ways computers learn from data.
Classification, Regression, and Clustering explained visually
Compare category prediction, number prediction, and discovering natural groups.
Classification predicts a category
Classification answers a which-kind question: dog or zebra, normal email or spam, benign or malignant. Supervised classification learns from examples that include the correct category label.
Regression predicts a number
Regression answers a how-much question. House price, temperature, and travel time are continuous numerical results rather than named categories.
Clustering discovers groups
Clustering searches for groups of similar examples without predefined labels. It can reveal structure based on characteristics such as color, shape, measurements, or behavior.
Begin with the question
Choose the task from the result you need. Use classification for a category, regression for a number, and clustering when you want to discover groups. Good machine learning begins with a clear question and appropriate data.
Quick check
Which task predicts spam or normal?
Classification.
Which task estimates a house price?
Regression.
Which task can begin without labels?
Clustering.
Lesson 6 transcript
Machine learning problems can ask different kinds of questions. Some ask which category an example belongs to. Others ask for a number. Some ask the computer to discover groups without receiving category names.
Classification predicts a category. A model can classify an image as a dog, zebra, or horse. It can classify an email as normal or spam. During supervised classification, the training examples include the correct labels.
Regression predicts a numerical value. Imagine using the size, location, and age of a house to estimate its price. The answer is not a category. It is a number that can change continuously. Weather temperature prediction and travel-time estimation are other regression examples.
Clustering finds groups of similar examples. Imagine placing a mixed collection of colored objects on a table without receiving instructions about their names. You may naturally group them by color, shape, or size. A clustering algorithm searches for similar structure in data without using predefined category labels. This makes clustering a form of unsupervised learning.
Classification and regression normally use labeled examples, so they are common supervised learning tasks. The correct outputs guide the training process. Clustering normally begins without those answers and searches for patterns within the inputs.
Choosing the task depends on the question. If you need to identify whether a tumor is benign or malignant, use classification. If you need to estimate a house price, use regression. If you need to discover groups of similar customers or patients, clustering may help.
These methods can work with many kinds of data, including images, text, measurements, and sound. The representation and model may change, but the task still begins with a clear question and appropriate data.
You have now completed the first part of Neural Networks for Everyone. You learned how artificial neurons combine inputs, weights, and bias. You saw how activation functions shape signals, how layers build complex features, and how feedback supports learning. You also compared classification, regression, and clustering.
In the next part of the Learning Center, we will explore computer vision and see how machines find information inside images and video.