INDEX STRUCTURE FOR SIMILARITY SEARCH OF NUMERICAL SERIES DATA, AND SIMILARITY SEARCH METHOD USING IT

摘要

PROBLEM TO BE SOLVED: To provide: algorithm for significantly improving a degree of accuracy in refining an approximate solution for the correct solution for an L-index and compensating an error of an approximate line segment for this; an index mechanism to be used for this; a similarity search method using it; and a medium for storing a program for implementing the method.;SOLUTION: The errors regarding all the possible cases are measured in advance and this maximal value is held as metadata. Degree of compensation is prevented from being excessive by compensating the error using this metadata. This enables the error to be compensated so that an approximate distance, or the distance of an approximate segments with each other, can be close to the real distance in the L-index. A problem of inefficiency, due to an excessive degree of this compensation, is solved and the approximate distance can be compensated closer to the real distance.;COPYRIGHT: (C)2004,JPO&NCIPI