Training data
The examples a machine-learning model learns from during development. Its coverage, quality, labelling and provenance largely determine what the model can do and where it will fail.
Why it matters
A model rarely performs better than its training data allows. Gaps in coverage, such as a region, customer type or season that is under-represented, often become the places where the model is least reliable in production.
Training data also carries legal and contractual obligations. Organisations need to know where it came from, whether they are permitted to use it for this purpose, and how it will be retained or deleted.
In practice
For example, a UK facilities management company building a model to predict HVAC faults might discover that its sensor history covers mostly newer buildings and add data from older sites before training, so predictions are not skewed towards modern equipment.
Where Rodan fits
Rodan assesses and prepares training data before model development in AI and Decision Systems delivery. See also how to get your data ready for AI in six practical steps.