Based on the research of Apriori algorithm of boolean matrix, an improved PMApriori(Prune Matrix Apriori) algorithm is proposed, and this algorithm could efficiently and speedily find frequent itemsets from the transaction database. For storage and computing characteristics of the matrixes are fully used, only once scanning of the database is required, and it is unnecessary to generate the candidate itemstts, thus the operation efficiency of the algorithm is improved greatly. The uninterrupted trimming and compressing of storage matrixes in the algorithm implementation could save the memory space and the matrix-scanning time. Demonstration and comparison of various algorithms indicate that the proposed algorithm has fairly good time and space characteristics.% 为了高效且快速地从事务数据库中生成频繁项集，在研究基于布尔矩阵 Apriori 算法的基础上，提出一种改进算法 PM Apriori(Prune Matrix Apriori)算法。此算法充分利用矩阵的存储与运算特性，仅扫描数据库一次，不需要生成候选集，有效地提高算法的运算效率，在算法执行过程不断对存储矩阵进行修剪压缩，节省了内存空间和扫描矩阵的时间，并通过算法示例与算法性能对比分析证明所提出的改进算法具有良好的时间特性和空间特性。
展开▼
