Association Rule Mining Algorithm

With the development of IT technology and popularization of Internet, a large number of data in the databases have been accumulated. At present, data mining has become one of the most active research fields for analyzing and understanding data, revealing knowledge that hides inside it. Mining association rule is one of the important research branches in data mining. Its purpose is to concentrate on finding the relation hidden between interesting attributes from the large-scale data.The content of this paper includes two respects: research of data mining algorithms and the system structure of the application platform of data mining. Seen from the present, the research of association rule algorithms have paid more attention to the improvement of efficiency, then to ignore the quality of the rules, which caused a lot of problems, such as huge and complicated rule sets, inaccurate information unable to understand etc. For this reason, this paper proposes to measure the association rules from certainty, utility, simplicity, interestingness and integrity. After pointing out the problems in traditional algorithms, this paper proposes a new approach to mining association rules based on comprehensive measures. This algorithm is based on classical Apriori algorithm and keeps minimum support and minimum confidence. The mining quality and efficiency have been greatly improved by introducing the concept of rule length and interestingness constraint and dividing the result rule set into positive and negative logic rules. In the aspect of the integrity of rule sets, the negative set and negative rule concept are put forward. The calculation method of support, confidence and interestingness of negative regular, which is given through three theorems, avoids low efficiency caused by repetitive scanning database. It is good supplement for rule sets. In the research of the application platform in data mining, DMAP, which is based on C/S structure, is designed. The system structure and functions of individual subsystems are given in detail. The platform integrates various kinds of data mining algorithms and models.According to different strategies and the goals, users can develop the solution schema for actual problems. The system has good generality, flexibility and expansibility and can offer effective decision supporting service for every professions and trades.