Research And Implementation Of Interactive And Visual Mining Of Association Rules

Visual Data Mining (VDM) is a data mining technology integrated with visualization. This combination break the close property of traditional data mining process, and make it possible to introduce and exert human knowledge and imagination into the process of data mining, and control the direction and process of data mining.Association Rules (AR) is one of the most widespread data mining tools. The purpose of Association Rules is to discover interesting rules between the itemsets in huge transaction database. However, traditional AR mining tools, evaluate rules by the framework of support-confidence frequently, cannot highlight the interesting rules. This is largely due to that user cannot exert his (her) domain knowledge in this problem. Therefore, this paper firstly do research on interest measurement of AR and propose an integrative interest measurement which contain objective and subjective interest measurement. Secondly, it focus on visualization of frequent itemsets, which is an important concept of AR. Consequetly, a visualization of frequent itemsets which embody the idea of Focus-plus-Context is put forward. Finally, on the basis of the Focus-plus-Context visualization, an interactive and visual AR mining model is designed. The model can unfold the close mining process, adjust the parameter flexibly, control the process, prune manually and highlight the interest. The proposed interactive and visual technology not only provides reliable and concise information by intuitive views, but also participates in the process of mining. Meanwhile, user's domain knowledge and purpose are embedded in data mining via abundant interactive means.In order to validate the model aforementioned, an interactive and visual AR mining subsystem (ivAssociation) is designed and implemented on the platform of MFDM (Multi-Functional Data Mining tool), which is an open and universal data mining tool. The results demonstrate that plugging visual and interactive technology into the process of AR is helpful for discovering interesting rules.