Efficient indexing methods for aggregation of temporal data

Abstract

Temporal databases (TDBs) manage time-evolving data. They provide built-in supports for efficient recording and querying of temporal data. Data in real world have temporal aspects and many applications, such as trend analysis, version management, and medical record management, handle temporal aspects of underlying data. So, DBMS should provide temporal support directly in these cases The TDB can be applied to applications like trend analyses, version management and video data management. The temporal aggregation in temporal databases is an extension of the conventional aggregation to include the time concept on the domain and the range of aggregation. The temporal aggregation is an important operation that is essential to many applications. There have been several proposals for the temporal aggregation processing. However, for various analytical applications in the TDB, it is not sufficient to consider only time attribute with the time attribute. We call the temporal aggregation t hat i ncludes more than one range-condition attribute a s w ell a s the time attribute in the condition the Multidimensional Temporal Aggregation (MTA). The MTA is very useful especially for large historical data warehouses. In this thesis, we propose a structure for the temporal aggregation, called the CTA-tree, and an aggregation processing method based on the CTA-tree. We perform experiments to show the effectiveness of the proposed method. The experimental results indicate that the CTA-tree works better than the other method. We also propose structures for the MTA, called the ITA-tree and the PTA-tree. Through analyses and performance experiments, we also compare the proposed structures with an intuitive extension of the SB-tree that was previously proposed for the temporal aggregation. The results show that proposed structures are superior to the intuitive extension of the SB-tree.