传统缓存算法存在命中率低、交换率高等问题,且现有缓存算法在分布式大数据存储系统中并不适用,为此提出了一种基于频繁序列挖掘的自适应缓存策略。该方法使用数据挖掘算法挖掘历史访问窗口内的频繁序列,将频繁序列模糊合并后构建匹配...传统缓存算法存在命中率低、交换率高等问题,且现有缓存算法在分布式大数据存储系统中并不适用,为此提出了一种基于频繁序列挖掘的自适应缓存策略。该方法使用数据挖掘算法挖掘历史访问窗口内的频繁序列,将频繁序列模糊合并后构建匹配模式集合以供查询。当新的访问来临时,将固定访问长度内的子序列与匹配模式集合进行匹配,然后根据匹配结果预取数据,同时结合修改后的S4LRU(4-segmented least recently used)数据结构进行缓存数据换出。在公开的大数据处理trace集上进行了仿真实验,实验结果表明,在不同的缓存大小下,提出算法与现有典型缓存算法相比,平均命中率提高了0.327倍,平均交换率降低了0.33倍,同时具有低开销和高时效的特点。此结果表明,该方法较传统替换算法而言是一个更为有效的缓存策略。展开更多
The classic sequential frequent pattern mining algorithms are based on a uniform mining support, either miss interesting patterns of low support or suffer from the bottleneck of pattern generation. In this thesis, we ...The classic sequential frequent pattern mining algorithms are based on a uniform mining support, either miss interesting patterns of low support or suffer from the bottleneck of pattern generation. In this thesis, we extend FP-growth to attack the problem of multi-level multi-dimensional sequential frequent pattern mining. The experimental result shows that our E-FP is more flexible at capturing desired knowledge than previous studies.展开更多
文摘传统缓存算法存在命中率低、交换率高等问题,且现有缓存算法在分布式大数据存储系统中并不适用,为此提出了一种基于频繁序列挖掘的自适应缓存策略。该方法使用数据挖掘算法挖掘历史访问窗口内的频繁序列,将频繁序列模糊合并后构建匹配模式集合以供查询。当新的访问来临时,将固定访问长度内的子序列与匹配模式集合进行匹配,然后根据匹配结果预取数据,同时结合修改后的S4LRU(4-segmented least recently used)数据结构进行缓存数据换出。在公开的大数据处理trace集上进行了仿真实验,实验结果表明,在不同的缓存大小下,提出算法与现有典型缓存算法相比,平均命中率提高了0.327倍,平均交换率降低了0.33倍,同时具有低开销和高时效的特点。此结果表明,该方法较传统替换算法而言是一个更为有效的缓存策略。
文摘The classic sequential frequent pattern mining algorithms are based on a uniform mining support, either miss interesting patterns of low support or suffer from the bottleneck of pattern generation. In this thesis, we extend FP-growth to attack the problem of multi-level multi-dimensional sequential frequent pattern mining. The experimental result shows that our E-FP is more flexible at capturing desired knowledge than previous studies.
