关联挖掘中的时效度研究

Time-validity in mining association rules

传统的关联挖掘算法,以支持度和置信度作为评价标准来衡量规则是否有价值。然而,这种模式不能体现出数据的时效敏感特性,如Web数据和长期积累数据。文中将首次建立一个全新的时基模型来重新估计数据规则的价值,并给出时效度(time validity)作为新的规则价值衡量标准。最后,给出了基于这个新的时基模型的一种新并行算法。这种算法使得我们在挖掘过程中使用增量挖掘,而且使得用户可以通过互操作来优化挖掘过程。

Data mining, also known as knowledge discovery in database, has been recognized as a new area for database research. This area can be defined as efficiently discovering interesting rules from large collections of data. As for association rules mining, traditionally speaking, support and confidence are the evaluation index to decide whether or not it is a strong association rule. Yet this ignores the prescription factor of time-sensitive data, such as Web data or long-date stored data. In this paper a new time base model was firstly built up to estimate the interest of the rules. A new interest-measure called (time-validity) for rules based on this new model was presented. A new parallel algorithm for this new time base model fundamentally different from the known algorithms was also presented. Those techniques allow for increasable mining and support more user-interaction in the optimized rule-mining process.