Dissertation > Excellent graduate degree dissertation topics show

Divide and conquer strategy in the application of association rule mining

Author: LinZhenMin
Tutor: YeYangMing
School: Xiamen University
Course: Computer Applications
Keywords: Association rules Frequent Itemsets Divide and conquer strategy Incremental updates Mark
CLC: TP311.12
Type: Master's thesis
Year: 2002
Downloads: 79
Quote: 1
Read: Download Dissertation

Abstract

Data mining association rule mining is an important direction of the field. It takes a large transaction database project sets the relationship between the use of the customer's buying behavior, help in developing sales strategies, classification design. Association rules are usually two metrics: confidence and support. Confidence describe the intensity of the rules, and support the usefulness of the rules described. Usually specified by the user minimum confidence and minimum support. We are interested in confidence and support were not less than the minimum confidence and minimum support of the strong association rules. Discovery of association rules must first dig out all frequent itemsets, and then by the frequent itemsets generated strong association rules. The second step in which the calculation is direct and frequent itemset mining on a large transactional database, you need to spend a lot of time, how to design an efficient and scalable data mining mining algorithms become a hot research topic. Wherein Apriori algorithm has been proposed is a more effective method. Apriori algorithm uses an iterative technique, first calculate the frequent 1 - itemsets L ₁, and then by the frequent k-itemsets L _k calculate the frequent (k 1) - Item Set L _{k 1}. By the L _k calculate L _{k 1} uses production - Test method: Generate a set of candidate itemsets C _{k 1}, on which the candidate itemset pruning, and then calculated for each candidate itemset support, determine the frequent itemsets. Time spent whole algorithm produced and tested by the two main parts. In the Apriori algorithm, based on the paper first presents a divide and conquer approach based on association rule mining algorithm DCM. DCM according to a certain strategy, the L _k split into three disjoint subsets: L _K ^{1 , L _K ^{2 , L _K ^{3 . Thus, most of the candidate set can be generated separately in the three sub-sets, thereby avoiding some of the connections between subsets; while most of the pruning may be separately confined separately on each subset search, to avoid entire L _k Search and time-consuming. To achieve this purpose, we introduce an auxiliary data structure, an increase of some frequent item sets flags left and right flag. Calculation of these two flags is straightforward, time spent is smaller. DCM algorithm which can effectively improve the efficiency of mining frequent itemsets. Since association rule mining is usually carried out in a large transaction database, generally need to spend a lot of time. Therefore, we not only need efficient algorithms for mining association rules, but also requires effective algorithms to maintain, update and manage. This paper also considered when the minimum support changes association rules update issues. We divide and conquer strategy is applied to update the ideological issues, the design of the algorithm DCIUA, both the use of divide and conquer approach has the advantage, while leveraging existing frequent item sets, and thus the algorithm is effective.}}}

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer software > Program design,software engineering > Programming > Data structures
© 2012 www.DissertationTopic.Net Mobile