Dissertation > Excellent graduate degree dissertation topics show

Construction Method for Training Data Set in Classification Algorithm of Support Vector Machines

Author: YuXu
Tutor: XieZhiQiang
School: Harbin University of Science and Technology
Course: Applied Computer Technology
Keywords: Statistical Learning Theory Support Vector Machine Gaussian distribution Vector Projection Virtual Sample
CLC: TP391.41
Type: Master's thesis
Year: 2009
Downloads: 295
Quote: 1
Read: Download Dissertation

Abstract


Support vector machine is Vapnik et al proposed a new machine learning method, which is based on statistical learning theory, with the optimal machine learning methods to solve the problem, reflecting the statistical learning theory of structural risk minimization thought. However, in practical applications, often need to deal with massive amounts of data or uneven small sample data, so how to improve support vector machines for complex data processing capabilities to enable support vector machines broader range of applications, as the current one research focus. In this paper, this problem is mainly made the following two aspects: first, the classification problem for the uneven distribution of training samples, and some very small number of training samples class problem, we propose a virtual sample based on Gaussian distribution generated Methods. This method is based on the Gaussian distribution theory, making the generated virtual samples to get a guarantee on rationality, in addition to the method can take advantage of a priori knowledge, for a variety of classification problems, so adaptability is enhanced. Conjunction with the presentation of a virtual sample generation method using support vector machine algorithm on a standard database of UCI datasets genus Iris (Iris data set) and the KDD CUP 99 intrusion detection data sets for simulation experiment results show that this method can make full use of the first acquired knowledge, to generate enough virtual label with a reasonable sample, effectively improve the classification accuracy. Second, support vector machines for massive data problem of slow learning algorithm runs through research support vector distribution characteristics, proposed an improved support vector based on vector projection pre-selection method. Linear separable case will be based on support vector vector projection in the center of the pre-selection method replaced by Fisher linear discriminant vector detection algorithm obtained optimal projection vectors, in the nonlinear separable case will be based on support vector vector projection preselection method in the approximate center of the feature space vector is changed to the true center of the feature space vector. By choosing a more reasonable projection vectors, and therefore the classification performance guarantee under the premise that the method can take advantage of a small number of more boundary vectors instead of the original samples for training, greatly reducing training samples, improve the speed of SVM training . This paper presents the simulation results also verified that the method is effective and feasible.

Related Dissertations

  1. Research on Automatic Detection Algorithm for Substructure Distress of Highway Pavement Based on SVM,U418.6
  2. Research on Autamatic Music Structrue Analysis,TN912.3
  3. Research on Transductive Support Vector Machine and Its Application in Image Retrieval,TP391.41
  4. Fault Diagnosis Method Based on Support Vector Machine,TP18
  5. Process Support Vector Machine and Its Application to Satellite Thermal Equilibrium Temperature Prediction,TP183
  6. Research for Infrared Image Target Identification and Tracking Technology,TP391.41
  7. Study on the Road Condition Monitoring Based on Vehicular 3D Acceleration Sensor,TP274
  8. Research of Diagnosing Cucumber Diseases Based on Hyperspectral Imaging,S436.421
  9. The Research on Intrusion Detection System Based on Machine Learning,TP393.08
  10. Dimensionality Reduction Methods for Gene Expression Data Base on SVM,TP181
  11. Prediction of Binding Affinity of Human Transporter Associated with Antigen Processing,R392.1
  12. Research on Feature Extraction, Selection and Classification Algorithms for Pulmonary CAD,TP391.41
  13. Research and Realization of License Plate Character Recognition Algorithm Based on SVM,TP391.41
  14. One kind of empirical data on the workload of a software bug fixes Prediction Model,TP311.53
  15. Based on support vector machine motion prediction image stabilization technology,TP751
  16. Based on self-learning social relation extraction research,TP391.1
  17. Based on Information Fusion Fault Diagnosis of Analog Circuits,TN710
  18. Multi-Class Pattern Analysis on Human Brain MRI Dataset,TP391.41
  19. Short-term Urban Traffic Forecasting Based on Multi-kernel SVM Model,U491.14
  20. Particle swarm optimization algorithm and support vector machine applied research,TP18
  21. Research on End-point Prediction Model of Electric Arc Furnace Based on SVM,TM924.4

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net  Mobile