经典的Apriori算法由于要多次扫描数据库，产生大量的候选项集，极大的影响了算法的时间和空间效率。为了减少数据库的扫描次数，可采用矩阵记录所需数据并放入内存中。矩阵record的每个行号对应一个商品码（即一个项），每个列号对应一个客户号（即一个事务）。倘若第j个客户买了第i个商品，则record[i][j]=1,否则record[i][j]=0. 扫描一遍数据库，将矩阵record初始化。接下来的工作，都可以通过扫描内存中的矩阵进行，不必再扫描数据库。(The classic Apriori algorithm generates many candidate itemsets because it needs to scan the database many times, which greatly affects the time and space efficiency of the algorithm. In order to reduce the number of database scanning, we can use matrix to record the required data and put it into memory. Every row number of matrix record corresponds to a commodity code, that is, one item, and each column number corresponds to a customer number (i.e. a transaction). If a j customer buys a I commodity, then record[i][j]=1, otherwise record[i][j]=0. scan the database once and initialize the matrix record. The next work can be done by scanning the matrix in memory without scanning the database again.)