Font Size: a A A

Research On Sequential Pattern Mining Based On Concept Lattice

Posted on:2006-12-28Degree:MasterType:Thesis
Country:ChinaCandidate:H J ZhouFull Text:PDF
GTID:2178360182956487Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Based on concept lattice which has been an effective tool for describing the hierarchical relationship between objects, researching sequential patterns mining which is an important data mining problem with broad application is a novel research point in data mining research field. Key algorithms and methods for mining sequential patterns based on concept lattice are deeply studied in this thesis, and the research results are applied in mining useful knowledge in undergraduate's marks database in universities. The main work and novel parts of this thesis are:1. Research on sequential patterns mining and concept lattice are in detail introduced, and basic methods and polices for constructing data structures and algorithms based on concept lattice for sequential patterns mining are analyzed.2. Batch algorithms and incremental algorithms for constructing concept lattice are studied. A New parallel constructing algorithm, called PIFGCL, which takes the advantage of the important properties of incremental constructing method and divides the constructing process into some parallel sub-process in order to achieve the parallelization of constructing process, is proposed and is implemented.3. Two kinds of concept lattice, IGCL made up of nodes constructed by itemset of transaction databases and SGCL made up of nodes constructed by sequential patterns of transaction databases, are proposed and are applied in mining sequential patterns. Policy based on IGCL for sequential patterns mining is proposed, and the process of mining sequential patterns is replaced by constructing process of SGCL.4. An evaluation system independent of courses for undergraduates' grade is proposed and some marks standards are proposed too based on it. Mark-Entropy is introduced to evaluate how much information is embedded in the marks of a course in order to estimate the importance of a course in the process of data mining.5. GMiner, a field data mining software designed for mining information from undergraduate marks databases, has been implemented.
Keywords/Search Tags:Concept Lattice, Sequential Patterns Mining, Data Mining, Undergraduate, Mark
PDF Full Text Request
Related items