Abstract: Six types of domain knowledge for data mining were summarized. A method of two-level knowledge organization, that is a combination of index knowledge and domain knowledge, was proposed. The former, as the knowledge catalog, codes the latter according to the characteristics of knowledge. Domain knowledge base consists of rule, function, hierarchical knowledge and basic information sub-bases. Syntax check for allkinds ofdomain knowledgewas discussed. Itwas pointed out that conflict forms of domain rule included conclusion conflic,t confidence conflict and inclusion conflict, and redundancy forms ofdomain rule included equivalence redundancy, inclusion redundancy and condition redundancy. Check algorithmswere presented subsequently. A knowledge base system, namedKB4DM (knowlege base for datamining), was implemented based on the research.