Skip to main content
Fig. 2 | BMC Bioinformatics

Fig. 2

From: Visually guided classification trees for analyzing chronic patients

Fig. 2

Data space partitioning by classification trees. The first step when creating a tree begins by considering the entire data space, represented in (a) for a two-dimensional dataset (with features X1 and X2), where at this point the tree only consists of a root node. The graphics also show samples belonging to three classes, which are coded as circles, squares and triangles. The first condition for splitting the data space is ‘ X1<30’, as shown in (b). In one region, all samples will have values for X1 less than 30, while in the other the values for X1 will be greater than or equal to 30. This creates two additional children nodes associated with each subregion, which can be further divided by considering mode conditions on the features. In (c), the two regions in (b) are split according to X2, generating two additional internal nodes and four leaves in the tree. Each leaf represents the most predominant class in their associated region

Back to article page