discrete classification

A discrete classification is where data items are allocated into a small finite set of classes such as cat/mouse/dog. These may be found in input data, or produced as a result of classification algorithms including machine learning. In some cases the discrete labels are known beforehand, in others generated by the analysis technique, for example using clustering algorthms.

Often some form of threshold is used to turn a numerical measure into a discrete classification. For example, we might chose critical values to classify a temperature as low/medium/high. Alternatively an algorithm might allocate a relative probability or confidence measure for each of several alternatives, in such cases one usually chooses the classification with the highest measure.