首先比较了现有的两种挖掘方法,提出了一种改进技术。综合考虑例外的局部和全局兴趣度,剔除非真正有趣的局部例外;增加两种客观度量并按模式重要度排序。实验表明该方法不仅可以有效挖掘多数据库中例外模式,而且还大大减少了用户负担。
The two existing techniques were compared and an approved method was proposed. The redundant patterns were eliminated through evaluating the local and globe interestingness. The exception patterns were ranked being added two interesting measures thus the user' s burthen could be reduced. Experiments on real datasets illustrate that the approach is efficient and promising.