For Heads-of · Practitioner
Data Quality
The degree to which data is accurate, complete, consistent and representative of what it is meant to measure.
- data quality
- data governance
- accuracy
- validation
Dimensions
Accuracy (correct values), completeness (no missing data), consistency (uniform formats and definitions), timeliness (current enough for the use) and representativeness (covering the populations the model will meet).
Impact
Poor data quality degrades a model directly: it produces unreliable output, reproduces whatever bias the data carries, and fails in production on the cases the data under-represented. Representativeness is the dimension most often missed, and its failure is training data bias.
Governance practice
Define quality standards for each dataset, check them before training and monitor them afterwards, and keep a record of where the data came from and what was done to it. Training data provenance checks carry that record.
Related Terms
Model Drift
A change in the relationship between a model's inputs and the environment it was fitted to, such that the model's assumptions no longer hold. The model itself is unchanged; the world it was trained on has moved.
Performance Metric
A quantitative measure of how well an AI model or system performs a task, such as accuracy, precision, recall, F1 score or a domain-specific measure.
Synthetic Data
Artificially generated data, created by algorithms or models, used instead of or alongside real data in training and testing.
Training Data
The dataset used to train a machine learning model, from which the model learns patterns and relationships.