| Data mining is boomed in the early 1990's. It's a key step in the process of knowledge discovery and is also a hot research topic in the domain of knowledge discovery. As an important task of data mining, association rule mining is applied to finding interesting association or correlation relationships among a large set of data items.In this paper, some techniques and theory about association rule mining are studied, and the efforts mainly focus on the improvement of mining algorithm and its application in education management decision. The main work of this paper is followed.(1) The algorithms of mining frequent items have been analyzed. The most classical algorithm for mining frequent items is Apriori. However, the vital flaw of Apriori algorithm is that multiple database scan is needed to add up the count of candidate item set members, which tempers its efficiency. To show it, the time spending is analyzed concretely in this paper.(2) An improved algorithm, Apriori-TIDS (Apriori Transaction Identifier Set) , is put forward in this paper, which can make up the above-mentioned flaw of Apriori very well. The point of Apriori-TIDS is that the sets of TID are used to add up the count of candidate item set members. The TID sets have been recorded while scanning the data base to generate candidate-1 sets. Thus, the count of candidate item set members is added up only by counting the number of corresponding TID set. The TID sets of the next rank of candidate sets are got only by intersecting the TID sets of the two frequent item sets which are used to be linked. The rest may be deduced by analogy, until all frequent item sets are found. Differing from Apriori, Apriori-TIDS needs only one database scan, and the count of candidate item set members is added up only by counting the number of corresponding TID set, except for the first time scan to create candidate 1-item set. Thus, the time spending is cut down greatly.The feasibility of this algorithm is demonstrated in this paper. According the characteristic of relationship database, Apriori-TIDS is programmed as a storageprocedure in ORACLE database to mine multiple dimension association rules. To be convenient for comparing, Apriori algorithm is also programmed as a storage procedure in the same database. The new algorithm is proved to have a better efficiency than the old one by experiments.(3) The basic framework of education information database and the association rule mining system are built. In according to the characters of education management information of Shanghai Maritime University (SMU), this paper builds the basic framework of education information database in ORACLE database, with the subject "Association Rule Mining in Education Management Decision" in the background, which is researched during my postgraduate period. Then, Apriori-TIDS is used to design and implement the ARMEMD (Association Rule Mining in Education Management Decision) system.(4) With the ARMEMD system, a lot of valuable association rules hidden in the education management information of SHMTU are found. Some of those rules are analyzed and the corresponding advices are put forward.(5) This paper also studies the management and visualization of association rules. To manage the rules, a rule template is introduced, by which the user can define interesting and uninteresting rule forms by himself, which can help the user find the rules he wants. As for the visualization, two methods, text and graph are implemented in this paper, which helps the user find interesting rules from large number of association rules quickly.Qu Chunjin (Computer Software and Theory) Directed by Prof. Yang Liping... |
PDF Full Text Request