Back to Blog

Understanding 3 Types of Machine Learning Algorithms


Machines learn through algorithms—sets of rules or instructions—that allow them to continually improve through experience. In contrast to programming, which can tell a machine to do a specific task in the same way over and over, machine learning algorithms modify themselves over time based on the original goal set and previous results. 

A simple real-life example of an algorithm is the creation or improvement of a recipe. You might follow basic sets of instructions or rules when following a recipe (e.g., preheating the oven or chopping vegetables), but have the freedom to make adjustments, like adding a little more salt or substituting one ingredient for another. The order of the steps is important, and not all substitutions are acceptable, but by making small modifications within a set of guidelines, you can improve the recipe over time.

Machine learning algorithms work in a similar way, but there are different types of learning styles depending on the outcome you’re seeking. As you don’t follow the same sets of rules when following a recipe, driving a car, or getting dressed in the morning, machines also need different rules for various types of tasks. Understanding how different types of machine learning algorithms, or models, work will help you choose the right one to achieve your desired outcome.

What Is Supervised Learning?

With supervised learning, the training data that the model uses to learn has a known label or outcome—there is a clear, expected answer. During the training process, the model makes predictions about the answer and is corrected when the result is incorrect. The model learns based on feedback, and accuracy improves over time until it has reached an acceptable level.

A recommendation engine for an e-commerce website is a real-life example of supervised learning. The model gathers data from previous images viewed and uses it to predict what new shoppers might buy based on their browsing habits. Every time a recommended product is added to the cart or a purchase is made, the model learns how accurate its predictions are and uses that information to improve its recommendations. 

Machine learning algorithms in this category include:

  • Artificial neural networks (ANN): Aim to mimic the human brain by learning to perform tasks based on provided examples

  • Bayesian logic (Naive Bayes): Classifies data based on the assumption that features are independent of each other 

  • Decision trees: Use simple decision rules to classify data based on predefined variables

  • Deep learning: Recognizes patterns based on training data, validation data, and test data

  • Linear discriminant analysis: Finds a linear combination of features to classify data

  • Linear regression: Predicts a dependent variable value based on a given independent variable

  • Logistic regression: Classifies binary variables such as pass/fail, yes/no, healthy/sick, and so on

  • Random forests: Use a combination of decision trees to deliver a mean prediction based on all of the trees

  • Similarity learning: Measures how similar two objects are to each other

  • Support vector machines (SVMs): Use data points in multiple dimensions to classify data based on a number of different features

  • Transfer learning: Uses information learned from previous data sets to inform decisions about related problems

Related: Download the checklist that lays out everything you should consider  before implementing a machine learning project or workflow →

What Is Unsupervised Learning?

With unsupervised learning, there is no clear answer and the model learns by deducing patterns or structures in a set of data that is not labeled. The learning is unsupervised because there is no human telling it whether the outcomes are right or wrong. It is up to the algorithm to identify the solution that makes the most sense based on the data. 

Unsupervised learning is like asking an open-ended question, whereas supervised learning is more like a multiple choice question that has a single answer. An unsupervised learning algorithm example in e-commerce is automatically segmenting customers into groups to provide different user experiences. The model will use the available data to determine what clusters make the most sense—product images viewed, product searches, product attributes, and so on.

Machine learning algorithms in this category include association, which is used to find patterns in large data sets with no labels, and those that use clustering:

  • Hierarchical clustering: Puts objects into distinct groups with other objects that are similar

  • Independent component analysis: Separates multiple variables into individual components

  • k-Means clustering, Association Rules: Partition observations into clusters

  • k-NN (k nearest neighbor): Classifies an object based on the similarity to its nearest neighbors

  • Principal component analysis: Creates a best-fitting line in multiple dimensions given a collection of data points

  • Singular value decomposition: Factorizes a matrix into singular vectors and values

What Is Semi-Supervised Learning?

Not surprisingly, semi-supervised learning is a combination of supervised and unsupervised learning. Some of the data is labeled and some is not. There is a desired outcome, but the learning process is not as simple as supervised learning. Because not all of the data is labeled, the model must also learn how to organize it.

One example of a data set that uses semi-supervised learning is image tagging for a large collection of images for an e-commerce website. Some images are tagged, others are not, and the machine has to learn how to classify the untagged images based on the existing labels and correctly group the images together.

Use Multiple Learning Types in One Platform

The beauty of the platform is that you can create different types of machine learning models and you don't have to be a data scientist to use it. However, it helps to have some understanding of the various types of machine learning algorithms so you know what is happening behind the scenes. uses supervised learning—specifically deep learning and transfer learning—to analyze data. This makes the platform well-suited to applications that require pattern recognition and those that require an extensive existing knowledge base, such as Natural Language Processing. To learn more about how to make your machine learning project successful, download our free checklist.