Coding indexing mode for time sequences

Abstract

The invention relates to a rapid indexing mode for carrying out arithmetic coding on variation trend of time sequences (determined time sequences and undetermined time sequences), which has important significance for predicting, classifying, carrying out data mining, carrying out knowledge discovery and the like on the time sequences. The indexing mode solves the problems of high data redundancy or low matching accuracy and indexing efficiency, which are caused by spatial indexing of the time sequences and can effectively complete similarity matching, clustering and classification of the time sequences (time sequences with equal lengths and time sequences with unequal lengths) by indexing and greatly reduce time complexity and spatial complexity of the problems of similarity matching, clustering and classification of the sequences. In the mode, firstly, time dimensions of the time sequences are subjected to interval segmentation, the time sequences are mapped into symbol sequences according to variation trend of the time sequences in each interval, then the symbol sequences are subjected to arithmetic coding to form unique code values and finally, the code values are used for establishing a B-Tree index for a time sequence set.

Description

Technical field [0001] The invention relates to a time series index method, which can effectively index the time series in a database, quickly retrieve the corresponding time series through the index, and effectively complete sequence similarity matching through the index. Background technique [0002] Due to the huge amount of data in the time series database, in order to quickly complete the retrieval and similarity matching, it is necessary to index the time series. Due to the high-dimensional nature of time series, and the Euclidean spatial distance is mostly used for the time series similarity measurement, most of the index methods also use the spatial index structure. From a large perspective, the spatial data index technology used in time series can be divided into two types: tree structure (including R tree, K-D tree, quad tree) and grid file, mainly including F-index, ST index, vp-tree , FastMap, etc., but these indexing methods cause high data redundancy or reduce...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the problems of high data redundancy and low indexing efficiency in the existing time series indexing methods, this invention proposes a coded indexing method based on time series trends. This indexing method can be used in O(n ) within the time complexity of encoding the time series into a unique code value, and then build a B-tree index on the time series through the code value, so as to quickly index the time series and effectively support similarity query matching through the index

Images