Glossary

What is machine learning?

Machine learning is a branch of AI in which a model learns patterns from data to make predictions or other outputs on new inputs. In product-data work it can support classification, matching and anomaly detection. Learning a pattern does not establish that every prediction is correct, so performance must be assessed for the intended task and data.

Patterns rather than a rule for every case

A conventional rule might assign a category when a particular exact condition is met. A trained model can use patterns across several inputs to predict a category or similarity. Supervised approaches learn from examples with known labels, while other approaches find structure in unlabelled data.

For example, a distributor could train or configure a model to propose product classes from descriptions and attributes. Staff can then review uncertain or consequential cases before the proposal becomes an approved classification.

Evaluate the task you actually need

A model can perform well on familiar products and poorly on a new supplier's terminology or a less common category. Keep a separate, representative evaluation set and inspect errors by product type, rather than relying only on a single overall score.

Classification accuracy, matching precision and missing-value detection are different measures. Choose the relevant outcome and account for the cost of false matches or incorrect classes. Deep learning is a family of machine-learning methods; generative models are one application area, not a synonym for every machine-learning task.

Related terms

Sources and further reading

AI-ready product data

Discuss the product information and controls your AI use case needs.