Machine learning is used in sentiment analysis to classify opinions, medical imaging to detect disease patterns, and robotics to support perception and action selection. In each case, a model learns from data rather than relying only on explicitly programmed rules.
How These Applications Work
A machine learning model is trained on data, tested with unseen data, and then used to make predictions or decisions. Inputs, outputs, and learning methods differ by application.
| Application | Machine learning mechanism | Example and limitation |
|---|---|---|
| Sentiment analysis | Natural language processing converts text into features that a classification model uses to label sentiment as positive, negative, or neutral. | A company may analyse reviews. Sarcasm, context, and biased training data can reduce accuracy. |
| Medical imaging | A supervised learning model trains on images labelled by specialists. Neural networks can identify patterns associated with tumours or other abnormalities. | The model may flag a suspicious scan for a doctor. False positives and false negatives require human oversight; patient data must be protected. |
| Robotics | Sensors provide environmental data. Computer vision identifies objects, while reinforcement learning can improve actions through rewards and penalties. | A warehouse robot may navigate obstacles and select packages. Unfamiliar conditions or faulty sensors can cause unsafe decisions. |
Effective systems need training data that represents operational conditions. Performance must be evaluated with suitable test data because accuracy alone may hide class imbalance or weak performance for particular groups.
A common misconception is that machine learning systems understand information as humans do. They identify statistical patterns, so outputs depend on data quality, model design, and evaluation.
IB Exam Technique
For an A4.1 response, identify input data, explain how the model learns or classifies patterns, and state the output. If asked to discuss or evaluate an application, give both a benefit and a limitation, such as efficiency versus bias, privacy, reliability, or safety.
