Efficient construction of near-optimal binary and multiway search trees

Abstract

We present a new linear-time algorithm for constructing multiway search trees with near-optimal search cost whose running time is independent of the size of the node in the tree. With the analysis of our construction method, we provide a new upper bound on the average search cost for multiway search trees that nearly matches the lower bound. In fact, it is tight for infinitely many probability distributions. This problem is well-studied in the literature for the case of binary search trees. Using our new construction method, we are able to provide the tightest upper bound on the average search cost for an optimal binary search tree. © 2009 Springer Berlin Heidelberg.