accuracies

Accuracy is a measure of how close a quantity is to its true value. In measurement and metrology, accuracy describes the closeness of agreement between a measured value and a true or accepted reference value. It is different from precision, which refers to the consistency of repeated measurements; an instrument can be precise without being accurate if it is systematically biased.

Accuracy is affected by systematic errors (bias) and random errors. In practice, accuracy is assessed through calibration, comparison against standards, and stated as a maximum error or an uncertainty interval. In formal metrology, accuracy includes both bias and dispersion of values within a given confidence level; traceability to national or international standards is often required.

In statistics and machine learning, accuracy is the proportion of correct predictions or classifications in a dataset. It is widely used but can be misleading on imbalanced datasets; alternative metrics include precision, recall, F1 score, and area under the ROC curve.

In navigation and geolocation, positional accuracy describes how close reported coordinates are to true positions; standards distinguish horizontal and vertical accuracy. Accuracy depends on signal quality, clocks, and processing, with typical specifications given as a distance (meters) at a confidence level.

In instrumentation, instrument accuracy is often specified as a percentage of reading or a fixed unit, accompanied by a stated resolution and uncertainty. Regular calibration reduces systematic errors.

Overall, accuracy is a context-dependent concept used across science and engineering to describe closeness to the true value, but it must be interpreted with its accompanying uncertainty and the presence of bias.