Supervised Machine Learning Algorithms

Name: Supervised Machine Learning Algorithms
Author: Briana Brownell

Discover applications for supervised machine learning including artificial neural networks

Preview the first lesson free — get full access to all 6 lessons.

Course: On-Demand

Supervise machine learning is the most common style of machine learning used because it solves many important problems of prediction and classification. Historically challenging problems like machine translation and image captioning are supervised machine learning problems because they have an output that can be correctly predicted. Unlike unsupervised machine learning, supervised machine learning relies on having a data set with an outcome variable. The outcome variable can be one of several types. You may want to identify the presence or absence of a feature. Examples might be identifying whether a specific financial transaction is fraudulent, determining whether a patient has a particular illness based on medical images, or predicting if a customer will cancel their service within the next year. This is called a binary outcome variable, which might be categorical. You may have several categories that you want to differentiate between. For example, you might want to identify the genre of a piece of music or the type of product in an online shop so it can be categorized. An outcome could be restricted to a single category or indicate the presence of a specific element. Finally, the outcome variable may also be continuous and have a range of numerical values. For example, you might be looking to predict the selling price of a house. Although the outcome variable is critically important to consider, without the right input variables, your prediction may not be accurate. Input variables are also sometimes called features or predictors. The terminology in machine learning is slightly different from traditional statistical modeling. In classical regression, input variables are usually called independent variables, and the outcome variable is called the dependent variable. The reason for the different terms is that several fields use predictive modeling and have developed their own techniques and terminology, including medicine, economics, sociology, and computer science. And although classical regression is an important part of predictive modeling in supervised machine learning, there are other algorithms used in classification prediction and forecasting. Artificial neural networks are frequently used in supervised machine learning. These techniques have found such a prominent place in the data scientist toolkit because the performance of artificial neural networks on some of the most challenging supervised machine learning problems has far exceeded the benchmark set by other methods. However, their use has brought new challenges. There's typically a trade-off between several decision factors for building a model, including accuracy, explainability, training time, and complexity. While neural network models tend to be quite accurate, explainability can be a complicated issue. That's because it's difficult to dig into the model to figure out why it's making the classification or prediction it is. These difficulties in explainability mean neural network-based models are harder to validate and can sometimes fail unexpectedly. In situations where the model output has a major impact, this trade-off might be unacceptable. Because of their complexity, training an artificial neural network on a large dataset with many features can take a large amount of computing resources. Although access to computing resources is getting cheaper and easier all the time, this means a data scientist has a while to wait before they can determine the success of the model they're building and troubleshoot any issues at each build. All these factors are trade-offs that a data scientist must make. If the model is too simplistic, it won't capture features relevant to the prediction task, but if the model is too complicated, then it might not fit the data and fail to generalize when it's deployed.

Intermediate Provider : Briana Brownell 6 Lessons · 21m in English

Course Description

Supervised machine learning is used in many every day applications, including image captioning, forecasting, and machine translation. But, because it’s so widely applicable, there are many different methods that can be used on different kinds of data to solve various styles of classification and prediction. These include regression, neural networks, and decision trees.

All machine learning heavily relies on data. In these lessons, we’ll consider some of the challenges in data collection and preprocessing that a practitioner should consider while building a machine-learning model. We will also review a number of examples of applications of supervised machine learning. You’ll also discover the reasons why supervised machine learning has been so successful, as well as some of the common challenges with using it. These lessons will discuss the most important applications of supervised machine learning including classification, prediction, and forecasting. You’ll also explore artificial neural networks, one of the most common methods used for supervised machine learning.

What You'll Learn

Differentiate between data suitable for supervised machine learning and unsupervised machine learning
Distinguish between styles of supervised machine learning such as classification and prediction
Explain classification modeling
Apply supervised learning to classification, prediction, and forecasting tasks
Explore artificial neural networks as a common method for supervised machine learning
Address challenges in data collection and preprocessing when building a machine-learning model

Key Takeaways

Supervised machine learning is used in many everyday applications, including image captioning, forecasting, and machine translation.
Different methods such as regression, neural networks, and decision trees can be applied to different kinds of data to solve various classification and prediction problems.
All machine learning heavily relies on data, so practitioners must consider challenges in data collection and preprocessing when building a model.
The most important applications of supervised machine learning include classification, prediction, and forecasting.
Artificial neural networks are one of the most common methods used for supervised machine learning.

Frequently Asked Questions

What topics does this course cover?

The course covers supervised machine learning, classification, prediction, forecasting, artificial neural networks, and feature engineering.

What methods of supervised machine learning are discussed?

The lessons discuss methods including regression, neural networks, and decision trees, with artificial neural networks highlighted as one of the most common methods used for supervised machine learning.

What skills will I gain from this course?

You will build skills in automated machine learning, machine learning, machine learning algorithms, machine learning methods, machine learning model training, and supervised learning.

What are some real-world applications of supervised machine learning covered here?

The course reviews applications such as image captioning, forecasting, and machine translation, and focuses on classification, prediction, and forecasting.

Does the course address working with data?

Yes. Because all machine learning heavily relies on data, the lessons consider challenges in data collection and preprocessing that a practitioner should consider while building a machine-learning model.