The Research On Multi-Datasource Classification Amalgamation Based On The Extending Of Concept Lattice

Classification has always been an important sub-branch of Data Mining. Nowadays, more and more Multi-Datasource data coming into people's daily life, and this also brings new challenge into Data Mining field, such as how to discovery knowledge from this kind of data and amalgamate them effectively. Therefore, the research on how to amalgamate classify knowledge efficiently from multi-datasource comes to be a important direction on this field.Concept Lattice is a model, which represents the knowledge with the relation between the intensions and the extension of concepts, and the relation between the generalization and the specialization between concepts. By introducing equivalent intension into Concept Lattice, we got the Extending Formal of Concept Lattice (ECL), which is an efficient tool to discovery classification rules. The content of the dissertation is as follows:1. All measures mentioned is this article were based on Extending Formal of Concept Lattice. Firstly, build ECL model respectively referring to each site;after that, extract knowledge from each model;finally, amalgamating them effectively. Two kinds of classify knowledge, classify rules and classify sub-lattice, were used here and their corresponded methods are improved both by theory and experiments.2. ECL has a high model complexity. On one hand, this feature could ensure the training set been classified very precision, on the other hand, it also often cause model Overfitting, which would hurt the performance of the classifier. So we take the model Pre-pruning technique into this article, and prevent the unnecessary sub-branch's appearance. Therefore, to reduce the complexity of the model and avoid the model's Overfitting.3. Based on the work stated above, a prototype system that can utilize multi-extending concept lattice classification knowledge discovery in database is implemented.