基于矩阵的多维关联规则算法在烟叶复烤配方的应用研究

Research and Application of Multi-dimensional Association Rule Algorithm Based on Matrix in Flue-cured Tobacco Formula

针对Apriori算法频繁扫描数据库造成的I/O开销大等问题以及烟叶复烤配方模块多维数据挖掘的需求,论文在经典Apriori算法的基础上,提出了一种基于矩阵的多维关联规则改进算法,该算法通过构造多维事务矩阵,减少了扫描数据库的次数,同时不断通过剪枝、剔除冗余事务对矩阵进行压缩,有效提高了挖掘效率。利用该算法对历史复烤配方数据进行了挖掘分析,结果表明该方法能够有效地将配方数据中隐含的诸多配方专家的配方维护行为规律提取为规则表示形式,从而挖掘出配方模块中烟叶的搭配和协同信息。该方法减少了配方研发人员的工作量,可有效地指导实际配方设计和维护工作。

This paper deals with the problem of high I/O cost caused by the frequent scanning database of Apriori algorithm and the demand for multidimensional data mining of flue-cured tobacco formula.Based on the classical Apriori algorithm,an improved matrix algorithm based on matrix is proposed.The algorithm reduces the number of scans of the database by constructing the multidimensional transaction matrix.At the same time,the algorithm compresses the matrix by pruning and eliminating redundant transactions,which effectively improves the mining efficiency.This algorithm is used to mine and analyze the data of the historical formula module.The results show that the proposed method can effectively extract the formula maintenance rules of the formula experts implied in the formula data into regular representation,and the collocation and cooperative information of the tobacco in formula module are founded out.This method reduces the workload of the formulators and can effectively guide the actual formulation design and maintenance.