Association Rule Mining In Concept Lattice Constructed Based On Rough Set Attribute Reduction

In recent years, as database management systems are widely used, and thedevelopment of computers and networks, which results in a large amount variety ofdata. Data Mining emerge as the times require, which causes a widespread concern ofinformation industry and the society, and becomes one of hot topics,in the pressingneeds of how to extract valuable knowledge and information from the mass data, so asto better use of these data to predict future trends.Association rule mining is one of important research contents and hot topics.Rough set and Concept lattice are powerful tools for data analysis and knowledgeprocessing. Rough set theory is a new mathematical tool, which mainly deals withincomplete and uncertain knowledge, and has been widely used in data mining andother fields. Based on attribute reduction in rough set is one of important researchcontents in the field of data mining, which can remove the redundant attributes,reduce the dimensions of the attributes, reduce the size of data mining, and thegeneration of a large number of candidate itemsets. While concept lattice wasgenerated formal concept analysis，each node of which is essentially a largest itemset.Calculate easily the support and confidence, and quickly find the interesting ormeaningful association rules by concept lattice constructed and the advantages ofHasse map visualization, so as to analysis the results of mining better, and obtainassociation rules better. Therefore, it is useful to extract association rules based onconcept lattice.In the paper, firstly we introduce the classical association rule mining algorithmsand analyse the studied problems further, including the generation of a large numberof candidate itemsets, scanning the databases many times, and low efficiency ofmining in the case of a larger number of data, and so on. Secondly, the classicattribute reduction algorithms is introduced, including Pawlak attribute reductionalgorithm, Entropy reduction algorithms and Skowron matrix reduction algorithms.However, these algorithms are either not intuitive, or more complex, or compute-intensive, or lower space performance, or difficult to achieve. Thirdly, as abreakthrough based on the problems, we integrate these two methods of the rough setattribute reduction and association rules based on concept lattice constructed so as toextract association rules. In the paper, the mainly research work and the results are asfollows:1) An improved decision table reduction algorithm is proposed from theperspective of using partition granularity in the rough set attribute reduction. In thisalgorithm, firstly, we define relative partition granularity and the relative importanceof the relative partition granularity, then using the relative importance as thetermination conditions, we obtain the relative attribute reduction set. The algorithm isa heuristic algorithm, the idea of which is that, firstly we obtain the relative corethrough the decision table, then based on the relative core, we find the relativereduction set, and analyzed and verified by experiment. The improved algorithm iseffective, and can be obtained accurate reduction sets, and has simple form, and easyto understand, and easy to implement on computer.2) In the paper, we integrate these two methods of the rough set attributereduction and association rules based on concept lattice constructed so as to extractassociation rules, the main idea of which is that, after the data pre-processing iscompleted, using the improved decision table reduction algorithm based onknowledge partition granularity deals with the completed data, in order to remove theredundant attributes, reduce the dimensions of the attributes, and the size of datamining, and the generation of a large number of candidate itemsets. Then, we get theconcept lattice of the reduction set by the algorithm of constructing concept lattice,which can visually perform the information of the decision table. Using the visualadvantage of the Hasse map, the users can easily calculate the support and confidence,and quickly find the interesting or meaningful association rules, so as to better analyzethe results of mining, and get better or interesting association rules. At last, theexperiment of global warming is analyzed and verified, which show that it is effectiveto integrate these two methods of the rough set attribute reduction and associationrules based on concept lattice constructed so as to extract association rules, and canget the better association rules.3) Based on the above studies, a prototype system is designed, which isAssociation Rules Mining bases on Concept Lattice Constructed in Rough SetAttribute Reduction. The prototype system has universal applicability, and the main advantage is that, it reduces the generation of a large number of candidate itemsets,and the size of data mining, and scan databases many times and the generation ofredundant rules, and improves efficiency and accuracy of data mining.