Research On Sequential Pattern Mining And Time Series Similarity Search

The types of data that data mining techniques are applied to include relational database, transactional database, spatial database, temporal and time-series database, world-wide web ,of which temporal and time-series data is one of the most important types on which our mainly research are sequential patterns and similarity search.This paper analyses all kinds of algorithms used on sequential pattern mining and discusses traditional similarity search techniques. Based on the above study, the major contributions of this paper include:1. Present a more efficient algorithm which uses the oriented graph to discover time-interval patternsTime-interval patterns is a novel research direction of sequential pattern mining. The algorithm that is introduced in this paper only needs to scan sequence database once, without using any candidate sequences. When sequence database is updated our algorithm only needs to scan the new added sequences to find new time interval patterns.2. Present a novel concept of "Threshold query".'Threshold queries' is to report those time series exceeding a user-defined query threshold at similar time frames compared to the query time series, it adds some constraint conditions related to attribute values in the similar searching process, so as to provide more satisfied and useful result for users.3. Present a faster algorithm based on threshold query.In this paper we let time interval sequence be stored in the form of discrete trajectories of multidimensional measurement points. At the same time, an efficient algorithm for calculating trajectory similarity is presented in the searching process to improve threshold queries' effect and efficient.