onehot

One-hot encoding is a method for converting categorical data into a numerical format suitable for machine learning. In one-hot encoding, each category is represented by a binary vector that has the same length as the number of categories in the feature. The vector contains exactly one '1' (the hot position) and all other elements are '0'. The position of the '1' identifies the category.

For example, for a feature with three categories — red, green, blue — red becomes [1, 0, 0], green [0, 1, 0], and blue [0, 0, 1].

One-hot encoded features are non-ordinal; the encoding does not imply any ordering between categories. They are widely used as input features for algorithms that require numerical input, such as linear models, tree-based methods, and neural networks. They are also used in natural language processing to represent words in a vocabulary as distinct tokens, and in classification tasks to denote the target class in a mutually exclusive setting.

Advantages of one-hot encoding include simplicity and little risk of introducing spurious ordinal relationships. Disadvantages include high dimensionality for features with large numbers of categories and sparsity, since most values are zeros. It also does not capture similarity between categories.

To address high cardinality, practitioners may use strategy variants such as embedding representations, target encoding, or hashing tricks. In statistical contexts, the approach is related to dummy coding or indicator variables. Many software packages provide one-hot encoding utilities, such as scikit-learn's OneHotEncoder or pandas get_dummies; it is important to handle unknown categories consistently between training and deployment.