delaycfim：基于滑动窗口的高速数据流闭合频繁模式挖掘方法word格式论文.docxVIP

优秀毕业论文 精品参考文献资料 以大大减少数据在线处理时间，从而可以对高速产生的数据进行有效的处理，避 免数据丢失的情况产生。 关键字：闭合频繁模式；流数据；滑动窗；数据挖掘 - III- ABSTRACT With the rapid development of information technology, data streams are generated in a variety type of applications. Hence data stream mining becomes a hot topic in data mining. Closed frequent itemset mining plays an essential role in data stream mining. It could be used in business decisions, basket analysis and network data mining, etc. Most methods for mining closed frequent itemsets store the streamlined information in compact data structure when data is generated. Whenever a query is submitted by user, it outputs all closed frequent itemsets. However, the online processing of existing approaches is so slow that those methods cannot deal with data streams generated at a high speed. The state of closed frequent itemset mining over data streams is analyzed in the thesis, and a novel method DELAY-CFIM for mining closed frequent itemsets is proposed to solve the problem of slow online processing. It divides the closed frequent itemset mining process over data streams into two steps. Firstly, when transactions are generated, it stores the frequency information of itemsets in a summary data structure. Then it mines closed frequent itemsets until a query is submitted. The method can improve the speed of online processing. The main contributions of this thesis are as follows: A summary data structure OTT is designed in the thesis, which is based on sliding window. OTT stores compact frequency information for the transactions in current sliding window, it compresses and aggregates information to reduce the cost of mining closed frequent itemsets. A new data structure for storing closed frequent itemsets called CFIT is proposed in the thesis. CFIT combines Tree with List structures. It stores closed frequent itemsets generated so far and helps to do the closure detection when necessary. The special structure of CFIT helps to improve the speed of closure detection. A n