Datapoint

Datapoint is a single unit of data that represents a measurement, observation, or fact. In statistics and data analysis, a datapoint typically consists of one or more attributes or features that describe an entity. In a tabular dataset, each row is commonly a datapoint, with columns representing features and, optionally, a target value. A datapoint may be a scalar value, such as a single temperature reading, or a vector of values, such as a multi-dimensional observation with several features.

In machine learning, a datapoint often comprises input features and, for supervised learning, a label or target. The term is sometimes used interchangeably with data point, though some writers reserve datapoint for an individual observation within a data structure.

Datapoints are the basic units for statistical analysis and visualization. They enable the computation of distributions, central tendency, variability, and relationships between features. The quality and representativeness of datapoints affect inferences and model performance; issues such as measurement error, sampling bias, missing values, and outliers can influence results. In large datasets, datapoints can be processed in bulk, sampled, or partitioned into training, validation, and test sets. In imaging, a datapoint can be a pixel value or a time-series observation, depending on context, though it is often treated as a feature within a higher-level datapoint.

Etymology: the term data point arises from data, the plural of datum; datapoint is a common compound form in informal contexts.