Font Size: a A A

Study And Design On The Algorithms Of Mining Association Rules

Posted on:2005-01-18Degree:MasterType:Thesis
Country:ChinaCandidate:W FengFull Text:PDF
GTID:2168360152955308Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the development of all kinds of network technologies and applications of Internet since 1990's, the growth of Internet traffic has been increased by the speed of exponent. In contrast to the rapid growth of information on the web, an effective means to store the knowledge, organize knowledge distributed on the Internet, and share knowledge is required urgently------to come into being the Data Mining(DM).The classical story, "Selling beers with urinary cloths", result in the powerful function of Mining Association Rules in Data Mining. Mining Association Rules will look for the correlations of the objects and refine it into the useful knowledge. The key task is for Mining Association Rules how to create the frequent sets. Based on the FP-Growth (Frequent Pattern Growth) algorithms, an improved algorithm is given in this thesis. Firstly, the business database is classified into some subgroups according to some rules. Then, the Frequent Pattern Tree is created and the Frequent Pattern is formed in the each subgroups. Finally, the improved results are obtained. On the other hand, the new algorithm is compared with FP-Growth in the performance. In Chapter 2, the basic concepts and the Apriori algorithm on Mining Association Rules is introduced. The Apriori algorithm requires to compute the candidates of the frequent pattern so that a lot of time is wasted in computing the candidate of the frequent pattern. In Chapter 3, the FP-Growth algorithm is analyzed and over the Apriori algorithm in the performance. For the algorithm, it is not require to compute the candidates of the frequent pattern. Based on the FP-Growth algorithm, a new algorithm is put forward and is compared with the FP-Growth algorithm. In succession Chapter, the new algorithm is proved that has some improvement on the above-mentioned algorithms by running the test program. At present, the Data Mining has become the hot-topic. The modified algorithm in this thesis is an only beginning on Data Mining, more works will be done in the future.
Keywords/Search Tags:Association Rule, Frequent Pattern, Frequent Pattern tree, the Candidates of the Frequent Pattern
PDF Full Text Request
Related items