Application Of Approximate Technology In Temporal Count And Sum Aggregation

Temporal database system is extended from traditional relational database system, therefore, the temporal database system should support temporal aggregation. For its increasing importance in the data warehouse applications, temporal aggregation query has become one of the most important fields of the temporal database research.Current temporal aggregation approaches have at least one of the following some shortcomings which are large space requirements, high processing cost and basing on complex data structures, which are not available in commercial products. In this paper, aiming at these shortcomings of current temporal aggregation algorithms, two improved algorithms basing on approximation techniques with bounded error are proposed. The first one uses the multi-versions B-tree to solve the problems about high level size consume. Since the application of approximate technology, this algorithms has logarithmic worst-case query cost. But the problem is this method is also basing on complex data structure, therefore it is impossible to implement in commercial products. To solve the disadvantages in current temporal aggregation, especially to solve the shortcoming about basing on complex data structure, the second algorithm merges B-tree and R-tree, these two data structures are widely used in current commercial products and the algorithm has achieved the same performance in the expected. The latter has overcome the most shortcomings of the current temporal aggregation algorithms thereby it is easy to implement in commercial products.Theories and experiment approaches are used in order to prove the validities of the two methods. The size consume of the two proposed methods, in this paper, are decreased more than an order of magnitude and query cost of them is greatly optimized. In the past, users retrieved data was based on exact query. In fact, what the users need was only a trend not an exact number, thereby the application of approximate technology can make users retrieve historical data more efficiently and know the development of the trend, through which users can decide the direction of their development.