Machine Learning (ML)

Machine learning is a type of artificial intelligence where systems learn from data rather than following explicitly written rules. Instead of a programmer telling the computer exactly what to do in every situation, the programmer gives the computer data and lets it figure out the patterns itself.

The fundamental idea

Traditional software follows rules: "If the customer's order is over £50, offer free shipping." A human writes every rule, and the software executes them precisely.

Machine learning flips this: "Here are 100,000 past orders and whether each customer came back. Find the patterns." The system discovers rules that a human might never think to write — perhaps customers who order between 7-9 PM on weekdays and buy two or more items have a 73% higher retention rate.

The three types of machine learning

1. Supervised learning: You provide labelled examples — input data paired with correct answers. "Here are 10,000 emails. These are spam, these are not. Learn to tell the difference." The vast majority of practical business ML uses supervised learning.

1. Unsupervised learning: You provide data without labels and let the system find structure on its own. "Here are 50,000 customer records. Find meaningful groups." This is used for customer segmentation, anomaly detection, and pattern discovery.

1. Reinforcement learning: The system learns by trial and error, receiving rewards for good outcomes and penalties for bad ones. This is how game-playing AI and robotics systems learn. It is less common in typical business applications.

Machine learning vs deep learning vs AI

These terms are nested: - All machine learning is a form of AI, but not all AI uses machine learning (some AI uses hand-coded rules). - Deep learning is a subset of machine learning that uses multi-layered neural networks. It excels at unstructured data like images, text, and audio. - Many effective ML systems use simpler algorithms — decision trees, linear regression, random forests — that are faster, cheaper, and more interpretable than deep learning.

Machine learning in everyday business

You are already using machine learning, even if you do not realise it:

• Email filtering: Your inbox separates important emails from promotions and spam using ML.
• Recommendation engines: Netflix, Amazon, and Spotify all use ML to suggest content.
• Fraud detection: Banks use ML to flag unusual transactions in real time.
• Search ranking: Google uses ML to determine which results are most relevant.
• Dynamic pricing: Airlines, hotels, and ride-sharing apps adjust prices using ML models.
• Predictive maintenance: Manufacturing companies use ML to predict equipment failures before they happen.

When to use machine learning

ML is the right approach when: - You have a large amount of historical data - The patterns are too complex for hand-written rules - The task involves classification, prediction, or recommendation - You need the system to improve over time as it processes more data

ML is not the right approach when: - You have very little data - The rules are simple and well-understood - Explainability is legally required and a black-box model is unacceptable - The problem changes so rapidly that models become outdated before they are useful