基于粒度计算的大数据集频繁项挖掘方法

Frequent Item Mining Method for Large Data Sets Based on Granular Computing

针对现有挖掘方法存在运行效率低下与精准度受限的问题,提出基于粒度计算的大数据集频繁项挖掘方法。通过分析大数据集中数据流的滑动窗口,得到支持数与频繁项之间的关联性,结合各种粒度计算模型,设计一种可以自适应不同种类粒度的计算方法,并采用六元组对其进行界定。运用能够反映数据特征的模式搜索比率,完成信息窗内结构粒的粒化,利用模式搜索比率的不同取值范围,获取模式搜索具备的属性特征,针对属性集的有序分类,嵌入所生成的子状态序列,通过得到的子状态压缩对象粒重构信息窗,使超级状态得以转换,最终取得频繁项集合。仿真结果表明,所提方法不仅能够提升挖掘精准度,而且缩短了运行时长。

Due to low efficiency and accuracy of traditional methods,this article puts forward a method of mining frequent items in big data set based on granularity calculation.By analyzing the sliding window of data flow in big dataset,we got the correlation between support number and frequent item.Combined with some granularity computing models,we designed a computing method that could adapt to different kinds of granularity,and then we used the six tuples to define it.Moreover,we used the pattern search ratio reflecting the data characteristics to achieve the granu-lation of structural particles in information window.According to different value ranges of pattern search ratio,we ob-tained the attribute characteristics of pattern search.Based on the orderly classification of attribute sets,the genera-ted substate sequence was embedded.Through the sub states,we compressed the object particles,and then recon-structed the information window,so that the super state could be transformed.Finally,we got the set of frequent i-tems.Simulation results show that the proposed method can not only improve the mining accuracy,but also shorten the running time.