What Is Machine Learning?

Date First Published: 18th November 2023

Topic: Computer Systems

Subtopic: Computer Software

Article Type: Computer Terms & Definitions

Difficulty: Medium

Difficulty Level: 6/10

CONTENTS

Uses Of Machine Learning
Types Of Machine Learning
Advantages and Disadvantages Of Machine Learning
History

Learn about what machine learning is in this article.

Machine learning is a type of AI that focuses on algorithms that can learn and adapt from historical data without explicitly being instructed. Machine learning algorithms use historical data to identify patterns and evolve their algorithms over time instead of following static rules. This enables the performance of AI to improve over time.

Uses Of Machine Learning

Uses of machine learning include:

Retail - Retailers use machine learning for personalisation, inventory management, and forecasting future demand. Machine learning algorithms can predict customer behaviour, product demand, and market trends by analysing historical data. This can help retailers make better decisions and adjust their strategies.
Online ad targeting - Google AdSense and Facebook advertising target ads towards users based on their past behaviours, such as what they have previously viewed.
Fraud detection - In industries like finance and ecommerce, machine learning algorithms can detect suspicious patterns of activity, like unusual transactions with debit or credit cards, or unusual logins, and help safeguard fraudulent activities in real time.
Dynamic pricing - Machine learning can allow businesses to adjust the prices they charge for products and services in real time based on changing market conditions, such as demand and consumer behaviour.

Types Of Machine Learning

The different types of machine learning include:

Supervised learning - Each piece of training data has one input and one label and the variables that need to be assessed for correlations are defined. Both the input and output of the algorithm are specified in supervised learning.
Unsupervised learning - Does not require data to be labelled. This means that its output is not provided to the system. Instead, these algorithms focus on finding patterns in unlabelled data.
Semi-supervised learning - Some training data is labelled and other data is unlabelled. The algorithm learns the dimensions of the data set using the labelled data, which it can then apply to new unlabeled data.
Reinforcement learning - Works by programming an algorithm with a specific goal and a pre-defined set of rules for achieving that goal. The algorithm is programmed to seek rewards for performing an action that helps achieve the goal and face negative feedback or penalties for bad actions. Reinforcement learning is often used in video gameplay and resource management.

Advantages and Disadvantages Of Machine Learning

The advantages of machine learning are:

It can more accurately identify trends and patterns to make predictions than following static rules, improving efficiency. The training of machines to learn from data and improve over time has allowed organisations to automate repetitive and detail-oriented tasks without breaks.
It enables businesses to gain detailed insights based on past data. This enables businesses to discover patterns, trends, and correlations that will be harder to manually collect.
It can provide a personalised experience for users. Machine learning algorithms are good at understanding user behaviour and preferences. This enables highly personalised suggestions to be made, such as targeted marketing campaigns and product recommendations.

The disadvantages of machine learning are:

It can be expensive to implement due to the complex software infrastructure.
Algorithms can be biased or poorly written. For example, algorithms may be trained on data sets that exclude certain populations. This can lead to unfair outcomes.
Since machine learning heavily relies on data to make predictions and learn from it, there can be privacy issues. Some owners may be able to peek at the data that is processed and sell it to third parties that will misuse it.

History

IBM engineer and pioneer of artificial intelligence Arthur Samuel introduced the term "machine learning" in 1959. The phrase was first used by him to describe a computer he created that could play checkers, evaluate past games, and make better play decisions in the future.