The Research And Application Of Association Rules Incremental Mining Algorithm

How to get different personalized information from a large number of data is the focus of information retrieval research. Research in this area includes meta-search engine and query expansion. The meta search engine which integrates multiple search engines return results focuses on the ability to provide users with more query results. And query expansion make search results closer to user demand through expanding the user-submitted short query to more keywords.Association rule mining is only an important data mining research direction. It is also an important method used in query expansion. This paper not only proposes proposes an improved association rule incremental mining algorithm but also proposes the concept of personalized meta-search engine with combining meta-search engine and association rules based on this association rule incremental mining algorithm.Firstly the association rule incremental mining algorithm used by query expansion has been argued. This paper has discussed factors that affect the efficiency of incremental mining on the FP-Tree structure and the strategy to fast update FP-Tree in FUFP. This paper has proposed fast update TD-FP-Tree algorithm(PFU-TDFP) with introducing FUP which is ne typical incremental mining algorithm based on Apriori to the TD-FP-Tree. The algorithm categorizes all items to reduce the possibility and times to scan the original transaction database. So does the number of reordering-items transactions to be dealt with when there are items become frequent from non-frequent. The algorithm further improves the efficiency by using parallel processing in some steps. The experiment shows that the algorithm proposed by this paper can not only fast update the TD-FP-Tree but also further improve the overall efficiency of mining in comparison with the structure of FP-Tree-based incremental mining.Then PFU-TDFP algorithm has been used to mine user’s search results browsing habits to make query key words reflect user’s industry background and interest tendency. The meta search engine is combined to propose the concept of personalized meta-search engine. A creative results fusion rated model has been proposed based on the local similarity such as the ranking of search results, the title and the summary for the results fusion of meta search engine. Eventually entire system prototype has been implemented, and the experiment for the system shows that the application of PFU-TDFP can fast incremental mine user search browsing habits. The Meta search engine results fusion rated formula proposed in this paper can provide personalized search results to users under the P@N method.