Unsupervised Learning

Unsupervised learning is a class of machine learning that finds structure in unlabelled data — clusters, patterns or outliers — without being told the correct answers. It is widely used in industry for anomaly detection where labelled failure data is scarce.

Because labelled examples of every fault are rarely available in industrial settings, unsupervised methods are valuable: they learn the normal pattern of operation from historical data and flag deviations from it. Techniques include clustering, dimensionality reduction and autoencoders. The trade-off is that without labels the model identifies that something is unusual but not necessarily what it means, so its outputs often need engineering interpretation.

Related terms

Supervised Learning · Anomaly Detection · Machine Learning (Industrial)