The Study On Privacy-preserving Association Rules Mining In Business Intelligence

With the rapid development of net commerce, more and more enterprises have urgent need of collecting and analyzing data efficiently, accurately and safely, and excavating the potential business opportunities, to defeat their opponent by a surprise move in the fierce competition. Facing massive scattered distribution of the data that can not be disposed centrally and enterprises'urgent need of central use, business intelligence (BI) thus came into being.The latest research of business intelligence is analyzed, and an in-depth study aiming at mining association rules in business intelligence, especially at the multi-level association rules mining and privacy preserving association rules mining is carried out, in order to improve the accuracy and security of analysis and decision in business intelligence. The main context of this article is as follows:(1) The issues of security and intelligence in business intelligence systems are carried out in-depth analysis; the latest research of data mining in business intelligence and privacy preserving association rules mining in business intelligence is reviewed.(2) The related concepts, architecture, workflow, etc of business intelligence are introduced, the effects of business intelligence to customer relationship management and the key technologies of business intelligence, especially the data mining theory and technology are analyzed.(3) The basic concepts, properties and the classic Apriori algorithm of association rules, especially the positive and negative association rules and multi-level association rules are introduced. For the issue that positive and negative association rules can excavate low-frequency, strong rules as well as many redundant uninteresting rules, we put positive and negative association rules and multiple minimum supports together to propose a method of positive and negative association rules based on multiple minimum supports. The method is verified by theoretical analysis and practical application that can excavate much significant negative correlation information as well as eliminate some insignificant rules and search apace. To the problem of mining cross-layer rules in multi-level association rules, we put granularity into multi-level association rules to propose a simple method of mining multi-level association rules based on granule compution. The method is verified by theoretical analysis and practical applications that can not only excavate the low-frequency and high-frequency rules between the low and the high levels, but also eliminate some redundant sub-rules in the"ancestor"rules.(4) The current privacy protection technologies, especially the privacy protection association rule mining technologies are introduced. For the privacy preserving problem that arising from the joint analysis of multi-databases in business intelligence, we applied secure multiparty computation into mining multi-level association rules, and offered the example analysis,simulation analysis and theoretical proof.