Dissertation > Excellent graduate degree dissertation topics show

Study of Ensemble Learning Based on Selective Strategy

Author: HanYanXia
Tutor: LiKai
School: Hebei University
Course: Computer Software and Theory
Keywords: Decision Tree Neural Networks Integrated learning Selective Ensemble
CLC: TP181
Type: Master's thesis
Year: 2011
Downloads: 39
Quote: 0
Read: Download Dissertation

Abstract


Integrated learning is one of the hot research in the field of machine learning. It is through some combination of multiple learning classification results fusion, thereby enhancing the integration of learning generalization ability to avoid over-fitting phenomenon. Integrated learning very wide range of applications, the main application areas of image recognition, voice recognition, classification of seismic waves. Selective integration is developed on the basis of the integrated learning, its main idea is a way to select part of the base model integration. This paper focuses on the selection strategy of integrated learning. The study mainly include the following aspects: first, using the ID3 algorithm and BP neural network algorithm to train a number of the base model. Here when using ID3 algorithm training base model the discrete processing data, using BP neural network algorithm to the part of the data were normalized, the processing of data affect the integrated performance; second climbing method through the new selection criteria, the preamble selection and subsequent three selective selective ensemble experiments proved the effectiveness of several methods, selection criteria parameters change also shows Integrated Performance differences produced a certain influence, and the experimental results the same for all base model integrated results compared the effectiveness of selective integration; Third, some clustering method of selective integration study, where clustering is of course to meet certain conditions, the base model poly to a cluster selection methods including hierarchical clustering and k-means clustering. Integrated model which clustering choice of four methods to select the center of the object as an integrated model, randomly selected from each cluster as an integrated model of an object, select the two objects as an integrated model, as well as randomly selected three objects as an integrated model Finally, a measure of the difference of these integrated model, metrics fail / no-fail, DF, and correlation coefficient method; generalization error analysis. Comparison experiments also verified the effectiveness of the selected center the validity of the object as well as selective ensemble. Which illustrate the selective strategy can improve the generalization capability of the integrated learning. The selective integration presumably in the future may also find their way.

Related Dissertations

  1. Designs and Applications of Fuzzy Synthetic Evaluation Models Based on Parallel Algorithms,TP18
  2. Research of QoS Optimization Based on Neural Network Prediction for Ethernet Passive Optical Network,TN929.1
  3. Researches on the Application of the Neuron-MOS in the Neural Network,TP183
  4. The Study of Intelligent Intrusion Detection System Based on Neural Network in Linux,TP393.08
  5. RFID Information Sharing in Supply Chain and Business Valuation Research,F224
  6. Study on Presentation and Implementation Strategy about Chinese Comprehensive Learning in Junior Middle School,G633.3
  7. Helicopter main gearbox fault diagnosis and failure prediction technology research,V267
  8. Passive analysis of the types of neural networks with delay,TP13
  9. Research on Predictive Framework of the Signal Control System,TP273
  10. The Index System of Risk Warning and Evealuation for Zhangshang Enterprise,F279.27
  11. Dynamic Reliability Evaluation of Existing Prestressed Concrete Bridge Based on Bearing Capacity,U441
  12. Missing Data Imputation in Bridge Health Monitoring System Base on Hybrid Model of SARIMA and Neural Network,TP183
  13. Research on the Design Features and Teaching Strategy in Chinese Integrative Learning,G633.3
  14. Neural network predictive control applied research,TP13
  15. Spam messages based on ensemble learning multi-level classification technology research,TN929.53
  16. Based on Artificial Neural Network Classification of Remote Sensing Images,P237
  17. Comprehensive Research on Intellective Deformation Analysis Method,P209
  18. Asymptotic Behavior for Some Kinds of Functional Differential Equations,O175.12
  19. Chinese Comprehensive Study,G633.3
  20. Ensemble Learning Study and Its Application,TP181
  21. A Study on the Support Vector Machine Ensemble Learning Mehtod Based on Particle Swarm Optimization,TP181

CLC: > Industrial Technology > Automation technology,computer technology > Automated basic theory > Artificial intelligence theory > Automated reasoning,machine learning
© 2012 www.DissertationTopic.Net  Mobile