Dissertation > Excellent graduate degree dissertation topics show

Research on Automatic Selection Methods of Data Mining Models Based on MAS

Author: GaoYaTian
Tutor: LiChunSheng
School: Northeast University of Petroleum
Course: Petroleum Engineering Computing Technology
Keywords: Data Mining Model MAS Knowledge-driven Agent Fracturing
CLC: TP311.13
Type: PhD thesis
Year: 2011
Downloads: 234
Quote: 0
Read: Download Dissertation

Abstract

Various sectors of the industry has accumulated a large amount of business data, there is an urgent need to convert these data into useful information and knowledge, and data mining (Data Mining, DM) caused great concern of the information industry. Data mining techniques to solve a wide variety of practical problems, the selection and design of the data mining model is a major part, is also the ability to deal effectively with the application of the key. Traditional data mining model the design dependent modeling expertise of the staff, to establish a mining model based on the analysis of the operational characteristics of the application areas of duplication of data exploration and algorithm testing, greatly reducing the efficiency of the model accurately. With the emergence of new technologies, model designers may have overlooked some important mining method and algorithm technology help knowledge discovery. Automatic modeling method of artificial low efficiency of modeling, knowledge reuse difficult problem for data mining, research data mining application features, technical characteristics and operational characteristics of the data, explore data mining models, design data mining model evaluation system; in the data mining Automatic Modeling Method based on the combination of MAS (Multi-Agent System) technology, the establishment of the framework automatically selects the data mining model based on MAS, and used in the field of oil field development. First of all, through the introduction of the mining model is automatically selected modeling concepts involved, defining feature, the basic meaning of the framework, objectives, activities, methods, entity developed data mining models to choose generic technology. Complete data characteristics, business characteristics, data mining technology features abstract definition to establish features of the system in the form of symbols; node-based mode, with target analysis, campaign analysis, designed automatically selects the framework of the three aspects of data mining models. The overall goal of the selection and design of the data mining models will be digging behavior abstraction for data preprocessing, preliminary model design, model adjustment, model evaluation, knowledge representation five basic activities. Data mining model selection framework specification data mining models automatically design objectives and activities of the various stages of, Organization mining business and mining technical features specific logical relations, the basic concepts of the design of the framework is based on the objectives, activities, methods, and the use of node expression mining process expectations, the response and measures required under different scenarios, expressed mining model to select different stages, at different levels of model design activities. In order to complete the a workable data mining model collection solving tasks designed based DMMS_F (Data Mining Model Selection based on Feature, DMMS_F) and DMMS_E (Data Mining Model Selection based on Experience, DMMS_E,) solving methods. Secondly, the design of the data mining model to evaluate the architecture, to study relatively applicable mining models from the collection of possible data mining model assessment method, mining model evaluation target modeling, and the completion of the model to evaluate the target specification describes mining model automatic selection mechanism; explore data mining model of comprehensive evaluation methods, including the design of the evaluation framework and evaluation factor. Taking into account the subjective factors and objective factors impact on the mining model evaluation, research, the adjustable design evaluation factor level position and the weight given data mining model evaluation system design the mining model quality evaluation method based on the level of the evaluation framework. Subsequently, the the MAS technology introduced data mining study of automatic model selection method, the establishment of MAS DMMAS (Data Mining Model Auto Selection) model-based framework; Agent cluster concept and design, through diplomatic role, management and labor roles realization Agent cluster mining model to select the design process of collaboration and interaction; research data mining model chosen to support the environment, reasoning with the running of the separation of the data mining model design, build a potentially effective data mining programs can be achieved on the basis of rational organization of knowledge selection and configuration design platform; Agent technology to explore the the the Agent architecture design and Agent collaborative model design; logical ring organizational structure based on the to design Agent dynamic management platform to achieve the dynamic management of the Agent, including the Agent dynamic management platform ring organizational structure and dynamic management platform for the management methods of the Agent; model based on analyzed from the perspective of system development based the MAS's DMMAS system design. Finally, exploring in the field of oil field development and production of oil well logging lithologic identification mining model selection framework and fracturing measures well selection system based on of MAS DMMAS applications, design, and evaluate its quality from the run and the application point of view. In the mining model design and application of data mining model in line with the characteristics of the universal choice modeling system, and give a specific model selection process and results from the development point of view shows the business definition, automatic data selection, mining model design, model comparison The data mining model is automatically selected to achieve the to instantiate completed based of the MAS DMMAS model.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer software > Program design,software engineering > Programming > Database theory and systems
© 2012 www.DissertationTopic.Net Mobile