基于模式矩阵的FP-growth改进算法

数据挖掘中关联挖掘算法比较典型的有Apriori和FP—growth算法．实验和研究证明FP—growth算法优于Apriori算法．但是针对大型数据库这两种算法都存在着较大缺陷，不仅要两次或多次扫描数据库，而且很难处理支持度和数据变化等关联规则更新问题．作者提出了基于模式矩阵的FP—growth改进算法，它至多扫描数据库一次，特别在更新问题上不用重新扫描数据库．通过实验结果分析，验证了这种改进算法相对于原有FP—growth算法的优势，特别在大数据集下，大大降低了挖掘的时间复杂度．

An Ameliorating FP-growth Algorithm Based on Patterns-matrix

The discovery of association rules is an very important aspect in data mining.There are Apriori and FP-growth algorithms among mining association rules algorithms.It has been proven that FP-growth algorithm is better than Apriori algorithm.But for very large databases there exists some big deficiencies in both algorithms,because two or more scans for the databases have to be done.It is also difficult to handle updating association rules in the cases including modifying support and inserting new data into the database.So,in the present paper,an ameliorating FP-growth algorithm based on patterns-matrix is presented which scans at most one for the database.Especially in updating problem,it needn't scan the database again.It indicates that the ameliorating algorithm is better than FP-growth algorithm through experience.The ameliorating algorithm sharply reduces the time come while mining those big datasets.