Soft decision trees are a variant of classical decision trees that use fuzzy set theory. They allow for fuzzy attribute values and classifications rather than the crisp (definite) values used in classical decision trees. Soft decision trees split nodes into overlapping fuzzy subsets rather than disjoint crisp subsets. They can match test examples to multiple leaf nodes and aggregate their outputs, rather than a single match as in crisp trees. This makes soft decision trees more suitable for databases with fuzzy or noisy attributes.