6.1 Automated and emerging technologies

Show All Section Notes

Machine Learning

1. What is Machine Learning?

Machine Learning is a method of data analysis that automates analytical model building. It is based on the idea that systems can learn from data, identify patterns, and make decisions with minimal human intervention.

2. ML vs. Expert Systems: The Core Difference

To understand why ML is powerful, we must compare it to the "Old Way" (Expert Systems).

Expert Systems

Rule-Based: Humans must manually write every "IF-THEN" rule.

Static: It only knows what it is told. It cannot handle a situation that wasn't programmed into its Knowledge Base.

Machine Learning

Data-Driven: The computer analyzes millions of examples to find its own rules.

Dynamic: It continuously improves. The more data it sees, the more accurate its predictions become.

3. Why ML Solves Problems Better

Machine Learning is superior for complex problems where humans cannot easily explain the rules. These include:

  • Unstructured Data: Recognizing a face or a voice involves billions of tiny variations. A human cannot write enough "IF" statements to cover every angle of a face.
  • Constant Change: Email spam changes every day. ML adapts to new spam patterns automatically, whereas an Expert System would need a human to constantly update the rules.
  • Hidden Patterns: ML can find connections in massive datasets (like DNA or financial markets) that are invisible to the human eye.

4. Worked Example: Email Spam Detection

This is a classic problem that Expert Systems failed at, but Machine Learning solved.

How ML solves Spam:

  • Data Collection: The system is fed millions of emails marked as "Spam" and "Not Spam."
  • Pattern Discovery: The ML algorithm notices that spam often contains specific word combinations, unusual sender IP addresses, and hidden tracking links.
  • Model Building: It creates a mathematical model of what a "Spam" email looks like.
  • Prediction: When a new email arrives, the ML model calculates the probability of it being spam.
  • Self-Correction: If you mark a "clean" email as spam, the model updates itself to be more accurate next time.

5. The Training Process

Machine Learning typically follows three stages:

  1. Training: Inputting massive datasets (Big Data).
  2. Testing: Giving the AI data it hasn't seen yet to check its accuracy.
  3. Application: Using the trained model in the real world.
⚠️ Exam Alert: If asked for the difference, remember:
Expert Systems = Human-made rules.
Machine Learning = Patterns discovered from data.