A data stream is used for handling dynamic databases, in which data can arrive continuously without limit. Association rule mining is a data mining technique, used to find the association between the data items in the databases. To generate association rules, frequent items are to be identified from the transactional database. Normally, in data mining, frequent-item-generation algorithms scan the database multiple times. But this is impossible in data streams because it handles dynamic databases. Hence, there is a need to develop a new algorithm, which reduces the number of database scans. In this work, two new algorithms named Scan-Reduced Indexing and Matrix algorithm are proposed for generating frequent itemsets in data streams. Performances of both algorithms are compared based on the execution time and the number of frequent items generated. Experimental results show that the performance of the Scan-Reduced Indexing algorithm is more efficient than that of the Matrix algorithm.