Font Size: a A A

The Research Of Frequent Pattern Mining And Applications In Supply Chain Management

Posted on:2011-07-30Degree:DoctorType:Dissertation
Country:ChinaCandidate:X Y ZhengFull Text:PDF
GTID:1118360308454571Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Frequent pattern mining is an important element in data mining. The parallel mining algorithms are payed more and more attention because of the huge amount of data to be processing in mining tasks. But there are many problems in specific areas of applications should be resolved. This paper discuss some special issuse, such as finding frequent pattern with multiple minimum Supports and data mining on sparse data source, and propose some new algorithms for mining frequent pattern on a novel view-oriented distributed sharing memory system: VODCA. The applications of frequent pattern mining in the supply chain management of integrated remanufacturing / manufacturing system are considered.First of all, the CD algorithm has good performance on distributed sharing memory system because of less communications among processors, But has no effective pruning strategy. A new algorithm with transactions reduction based on CD is proposed in this paper. The data set are initially classified using hierarchical cluster into sub-sets with different items. Computing tasks are allocated using the dynamic task allocation strategy among the processors. At the kth scanning of database, transactions that contain no k-frequent itemsets are pruned. Because of different item sets contained in different data subset, pruning plays role on some data subset.Secondly, in order to mining more patterns with large length, a method for mining frequent patterns using a redundant FP-tree is proposed. Because that frequent itemsets can be found from FP-tree directly, different minim supports are applied when the length of patterns reducing to generate patterns with different support. Piecewise function, with simple, flexible features, is used as the constraint condition. User can specify the length of the smallest patterns and the range of support. That makes finding interesting long patterns easy.Third, the sparse data source is a common data source of data mining. This paper analyzes a variety of data structure used in frequent patterns mining, and designs a list structure for compressed storage of the sparse data source. An algorithm for mining frequent pattern from sparse data sources is proposed to achieve a highly efficient. The division and the use of views on VODCA are discussed in detail.Finally, in supply chain management of integrated remanufacturing / manufacturing system, the choice of collectors and the professional dismantling center location problem are two key issues. A system structure of collectors'performance evaluation is proposed. This paper discusses how the frequent pattern mining used in such applications and how to select key attributes. The supervised quantitative based onχ~2 statistics and the value mapping of the attributes are used to transform the data source into sparse data source. Relationships between attributes found by data mining can be used to provide guidance when determining the various types of evaluation indicators.
Keywords/Search Tags:Frequent Patterns Mining, view-oriented parallel programming, multi-support, sparse data, integrated remanufacture/manufacture system, supply chain management
PDF Full Text Request
Related items