Research And Application Of Association Rule Mining Algorithm Based On Background Knowledge

Along with the continuous development of observation means, the handling problem of mass observation data with LAMOST telescope as representative is increasingly sharp, so that the traditional artificial or semi-artificial data analysis approaches cannot satisfy the demand of astronomy. For astronomers and scientists are engaging in the brainpower of astronomical work, the correlation interrelation between the spectrum data characteristics and its physics chemistry qualities is very important. Association rule describes the relationships among items in data sets. Therefore, adopting association rule to depict the relation of celestial body spectrum data property can offer accurate, reliable and easy disposed by computer the spectrum knowledge for astronomers, so it is very valuable. With LAMOST as application background, this paper studies Association Rule mining algorithm that is based on background knowledge, and the major research work is as follows:Firstly, on the foundation of detailed analysis background knowledge facing association rules, this paper brings forward a technology of the background knowledge representation, which is based on the logic of predication. Then, the background knowledge facing association rules will be divided into heuristics backgroundknowledge and inductive background knowledge.Secondly, during association rules, pattern count price is too high and the efficiency of I/O is too low, so this paper brings forward an Association Rule mining algorithm that is based on background knowledge. This algorithm cut frequent itemsets that isn't inconsistent with restraint condition through heuristics background knowledge, and only generate frequent itemsets that interest users, so it not only reduces the counting costs of models, but also improves quality of mining data. At the same time, this algorithm obtains inductive background knowledge through sampling mining, so that it not only can generate different length frequent itemsets in every time database scanning course, but also can excavate all frequent itemsets in more few database scanning frequency.Thirdly, on the foundation of above-mentioned research accomplishment, using association rule as the way of analyzing star spectrum data and VC++, SQL Server 2000 as development tools ,an interrelation analysis system on star spectrum data is designed and realized, at the same time, its software architecture and function modules are given. In addition, the paper elaborates the following key techniques: the preprocessing to star spectrum data, sampling techniques, background knowledge templates and Association Rules mining. In the end, the running results of the system show that it is feasible and valuable to apply association rule to discuss existing or implying interrelations between the spectrum data characteristics and its physics chemistry qualities.To make a conclusion, this paper studies some contents that include the representation technology for the background knowledge facing relation rules, an Association Rule miningalgorithm that is based on background knowledge, as well as the correlation analysis technology of the spectrum data of celestial body. It is of great importance in theory and practice for improving the efficiency and quality of Association Rules mining and realizing the knowledge discovery of the mass spectrum data of celestial body.