Research Of The Distance-Based Quantitative Association Rules Algorithms

With the development of database and Internet technology, the volumes of data and information which can be obtained increase at the speed of exponent. Data mining comes up to solve the problem that how to distill the useful knowledge which can be used for decision supporting from masses of data. Analysis of association rules in data mining is an important method, which finds the interesting relations between items or attributes of database. Association rule mining aimed to find the associated relationship of a great deal of item sets in the database. At present, the association rules in quantitative association rule mining has become an important research direction in the field of data mining.In the background of quantitative association rule mining, this paper has done the distance-based quantitative association rules researches. Firstly, we introduced some theories and concepts of data mining and association rule mining, and common algorithms. On this basis, the major study of the Apriori algorithm's problems which used for mining quantitative association rules. Then, by introducing the thinking of mining quantitative association rule of algorithms at present, we analyzed emphatically the algorithm of the distance-based association rules(DRA), and some improvements were proposed: First, improved the clustering part, which clusters all attributes of data with the K-means and CADD, so that clustering results reflect better the relationship between data; Second, reduced the qualification of the definition of distance-based quantitative association rules, so that this algorithm used more easily; Third, using the radius of cluster for setting the parameter D0 makes more reasonable and understands simply. The algorithm was programmed by VC6.0, and achieved the visualizing data mining. Finally, in order to testify the validity of the improved algorithm, the geochemical survey data in a certain area in China and clinical data were mined and analyzed by the improved algorithm. The mining results consist with the actual domain knowledge.