论文部分内容阅读
提出一种基于静态IS-树的频繁模式挖掘有效算法IS-mine,并与经典的Apriori算法和FP-growth算法进行了实验比较。算法直接构造频繁项集,不进行Apriori算法采用的代价较高的候选集产生与测试操作。算法采用深度优先,模式增长的策略,挖掘任务只在一棵静态的IS-树上进行,避免了FP-growth算法所采用的代价较高的动态树的构建。针对不同特征数据集算法采取不同的过滤技术来缩小搜索空间。实验与理论分析表明,对于稠密和稀疏数据两类数据集,算法都具有较好的时空效率。
This paper proposes an efficient IS-mine algorithm for mining frequent patterns based on static IS-tree, and compares it with the classical Apriori algorithm and FP-growth algorithm. The algorithm directly constructs frequent itemsets, and does not perform the expensive candidate set generation and test operations adopted by Apriori algorithm. The algorithm adopts the strategy of depth first and pattern growth. The mining task is only performed on a static IS-tree, which avoids the construction of expensive and dynamic trees used by the FP-growth algorithm. Different feature data set algorithms adopt different filtering techniques to reduce the search space. Experiments and theoretical analyzes show that the proposed algorithm has better spatio-temporal efficiency for both the dense and sparse data sets.