Study On Methods Of Evaluation For The Association Rules

Knowledge Discovery in Databases (KDD) is a very active field among the subjects of the databases technique, the artificial intelligence, etc. Data Mining is an important step of KDD, can discovery knowledge from the databases. It can be understood by users and is interesting, valid and potentially useful.Association Rules Mining, an important branch of Data Mining is the strong' association rules that can satisfy all the minimum support threshold and the minimum confidence threshold. Mining the association rules often needs dealing with large scale data. As a result, the number of the rules is large and not all the consequences can really reflect the knowledge of realistic fields and meet the users' needs . Therefore, it is necessary to evaluate the results of the Association Rules Mining.The current evaluation methods about association rules mainly consist of the objective and subjective evaluation methods. In objective evaluation, there are mainly two evaluation criteria: support and confidence. In the actual application, it is not enough to evaluate the association rules only using support and confidence. So, it is necessary to improve the support-confidence model and increase the evaluation standard so as to mine the more satisfying and efficient association rules. This dissertation concludes the current objective and subjective evaluation method and also makes some research work is as follows:1. Analyzes and compares the traditional model of association rules evaluation and also discuss the existing problems.2.Presents concept of frequent model and comprehensive evaluation method combined the objective and subjective measurements, which enhances the evaluation about association rules.3.At the basis of setting up the fixed threshold about support, discuss the methods of setting up the variable support and comparatively analyze them.4.Discuss the methods of setting up the automatic threshold about support, enlarge the relative concepts, and improve some formulas in some relative documents.