Research On Time Series Similarity And Temporal Association Rules Discovering

With the development of computer software and hardware, the ability of generating and collecting data by information technology has improved greatly. As a new kind of data analysis and predict technique, data mining technology has attracted broadly attention in the recent years. As a very common type of the data sets, time series has been one of the focuses of the current data mining research.After surveying the major issues in data mining in and outside our country, problems on time series' representation, similarity searching, sequential pattern mining, incremental mining and temporal rules discovering are analyzed, works and contributions are listed as follows:1. This paper analyzes the piecewise linear representation of time series and gives a new type of measurement named pattern trend distance. Also, this paper proposes a new similarity searching algorithm by using this new measurement, which adopts a new "omitting overlap" optimistic technology to improve the efficiency.2. This paper studies a variety of time series sequential pattern algorithms. Because the closed sequences have good properties, this paper proposes a better efficiency algorithm based on closed sequences. This method avoids the generation of the large sets of candidates, improves the efficiency.3. The problem of incremental mining in large databases is discussed in this paper. A new incremental mining of closed sequential patterns algorithm called InClospan is brought up.4. The problem of discovering the temporal association rules in interval-based database, which using the newly obtained closed sequential pattern algorithm is discussed in this paper. We can give prediction to the data by using the temporal association rules.