Recent work on discretization of continuous-valued attributes in
learning decision trees has produced some positive results. This paper
adopts the idea of discretization of continuous-valued attributes
and applies it to instance-based learning (Aha, 1990; Aha, Kibler &
Albert, 1991).