Andmepunktide

Andmepunktide, in Estonian, translates to data points. In statistics and data science, a data point is a single observation recorded in a dataset. It may be a scalar value or a vector of multiple measurements, representing one instance across several variables. In tabular data, each row corresponds to a data point, while the columns represent features or attributes. If the dataset is labeled for supervised learning, some data points carry target labels; in unsupervised settings, data points are unlabeled.

Data points are analyzed in feature space, where each coordinate corresponds to a variable. Distances between data points—using metrics such as Euclidean or Manhattan distance—help identify similarities, clusters, or outliers. Visualizations like scatter plots display data points in two or three dimensions, while high-dimensional data often require dimensionality reduction techniques.

Data quality affects the usefulness of data points: measurement error, missing values, and outliers can bias results. Data collection, preprocessing, and cleaning often aim to improve the representativeness and reliability of data points.

In practice, data points serve as the fundamental units in modeling, hypothesis testing, and inference. They are the inputs to machine learning algorithms, statistical estimations, and data visualizations. The concept applies across domains, from scientific experiments to market research, where each data point captures a specific observation under defined conditions.