| It is a hotspot that the data mining of time serial model, classify rule, association rule in the data mining study currently. In this thesis, the thorough study of time serial model, classification rule and association rule is made. It is thought that the data mining is the multistage process of user' s center in this thesis. It is important that the data mining of multistage process should be study in the each phase. In this thought, the main work in this thesis as following:The way of the data mining is deeply studied and a model of data mining based on data warehouse is introduced. The model is combined with the strategy of data warehouse and OLAP technology, made use of concept hierarchy as background knowledge, and extended data in the database as user' s interesting metadata of the concept hierarchy, aggregation.Besides, relativity analyses is introduced in the process of data pretreatment in this thesis, thereby canceled the disrelated attribute of data mining assignment, reduced lots of data sets and improves the accuracy and efficiency of rules mined.To be dealed agaist extended data, this thesis has improved on and come true arithmetic of time sequence model, and amended conventional decision tree arithmetic, introduced the decision tree arithmetic for extended data, namely threshold value control approach. According to threshold value and concept hierarchy, threshold value control approach can set up the concise and statistic classification tree.At the same time, based on the theory of the concept lattice, this thesis introduces the arithmetic of mining association rules based on quantified concept lattice reduced by uncertainty coefficient.This thesis designs and accidencily carries out the prototype system of data mining. This tool is centered on the user, under the user' s control, and to be capable to effectively mine the rule of time sequence model and the classification rule and the association rule in the database or data warehouse.
|
PDF Full Text Request