Trie hashing analysis

Abstract

The author presents an analysis of trie hashing for alphanumerical keys. He proposes a variant that uses a binary code and an asymptotic analysis of the size of the index. This provides, for biased distribution, a computable formula that predicts the size of the index as a function of the frequencies of the characters and the transition frequencies between these characters. These results are confirmed by a simulation. The author considers a Markovian probabilistic method and uses the Mellin transform.<>