TIME-SERIES DATAMINING SYSTEM WITH ANNOTATED INFORMATION

摘要

The purpose of this study is a development of a flexible datamining system which exploits mainly over numerical time-series data. The first step of this task requires processing of data from various points of view. For this purpose we develop time-series manipulation language TQL, which can operate over patterns in time-series data. We show an outline of the language. Another point of this study is a use of annotated information, which is obtained semi-automatically from the Web space. In most cases data, on which we carry out datamining, have back ground information. We propose a mechanism to search related short text information, which is called annotations, against specified points or intervals of time-series data. We show a method that carries out datamining with collected annotations.