Dissertation > Excellent graduate degree dissertation topics show

Support vector data description and fraud in the financial statements of Recognition

Author: LiuZuo
Tutor: ZhuYuQuan；ChenGeng
School: Jiangsu University
Course: Applied Computer Technology
Keywords: Support Vector Data Description Multi-class classification Boundary optimization Incremental learning Identify fraud financial statements
CLC: TP391.4
Type: Master's thesis
Year: 2010
Downloads: 145
Quote: 1
Read: Download Dissertation

Abstract


Support vector data description based on statistical learning theory as a single classification method in solving finite sample, nonlinear and high dimensional pattern recognition problem of data has shown many unique advantages, has become another field of machine learning research hotspot. In support vector data description structure requires only one class of sample information will be applied to the financial statements fraud identification studies, can solve the problem of corrupt data is not readily available, to reduce investment risks, and enhance the transparency of accounting information and the promotion of the healthy development of the market have an important role. Therefore, in-depth research support vector data description will have a very high academic value and practical significance. Paper summarizes the support vector data description research situation, analyzes the advantages and disadvantages of existing methods. For fuzzy support vector data description of the calculation of membership problems in the nuclear space presents a degree of membership in the calculation method of the sample, thus achieving a kind of hierarchical fuzzy support vector data description algorithm. Based on support vector data description for the multi-class classification algorithm for overlapping areas discriminant strategy adopted by the lack of kernel space is proposed based on the relative density of multi-class support vector data description classification algorithm. For existing borders optimization algorithm fails to fully utilize the sample in the nuclear space distribution information, we propose a new boundary optimization algorithm. Incremental support vector data for an existing algorithm described problems, we propose a support vector data description incremental algorithm. On support vector data description of research results based on the design and implementation of a fraud recognition model financial statements. The main thesis work includes the following aspects: 1, support vector data description summarizes the research status, describes the basic problems of machine learning and statistical learning theory, and support vector data description for a detailed discussion. 2, we propose a hierarchical fuzzy support vector data description algorithm KHFSVDD. The algorithm first use of nuclear K-Means K the original problem into sub-problems; Then, the application of fuzzy support vector data description algorithm to generate a partial description of sub-problems; Finally, by combining the solutions to subproblems to build a global description of the original problem. 3, we propose a kernel space relative density of thought, and applied based on support vector data description of the multi-class classification algorithm to kernel space relative density of decision-making basis to determine the overlap region hypersphere class test sample . 4, we propose a boundary optimization algorithm based on super-ball near the boundary of the average density of the sample information, as well as the distance between the test sample and the center of the sphere, near the border of the test sample on the category to judge. 5, we propose a support vector data description incremental improvement algorithm in the analysis of the composition of support vector set, based on a dynamic manner selected could turn into support vector data for training, the training samples at about minus while retaining more data distribution information. 6 was constructed based on support vector data description fraud recognition model financial statements, including the initial description of the model, incremental descriptions and statements of detection modules.

Related Dissertations

  1. The Research and Implementation of the Fault Management System for Triple Play,TP315
  2. Research on Problems of Pumping-jack Diagnosis Based on Support Vector Machine,TP18
  3. Support Vector Data Description fault diagnosis,TH165.3
  4. Key Technologies of WEB Media Resources Intelligent Discovery,TP393.092
  5. Based on incremental improvements in the Bayesian question classification in the field of research,TP391.3
  6. Support Vector Machine and Its Application in Intrusion Detection,O212
  7. Research on Alloying Component Controlling Model for LF Steelmaking Based on HS-SVR Algorithm,TF703.8
  8. A Study on Algorithm for Classification Based on Support Vector Data Description,TP181
  9. Music Automatic Classification Based on Support Vector Machine,TP391.4
  10. Text based on SVM multi-class classification,TP391.1
  11. Research on Question Classifying of Chinese Question Answering System Based on Bayesian Classification,TP391.1
  12. Study on Techniques of Rough Set Based SVM Hierachical Categorization,TP391.1
  13. A Study on Techniques of Hierarchy Classification and Incremental Learning,TP391.1
  14. Design of Bank Anti-money Laundering System Based on Anomaly Knowledge Discovery and Incremental Learning,TP311.13
  15. Research on Machine Vision Based Injection Product Defects Detection System,TP391.41
  16. Study of Fault Diagnosis of Analog Circuit Based on Combinating Support Vector Machine and Genetic Algorithm,TN710
  17. An Incremental Learning Method for Hierarchical Latent Class Models,TP18
  18. The Key Techologies of Fuzzy Support Vector Machine,TP181
  19. The Research on P2P Flow Detection Technology Based on SVM,TP393.02
  20. PCA - based dynamic facial feature extraction and its incremental learning algorithm,TP391.41
  21. On Ordinal Regression Machine and Multi-class Classication Algorithm,TP18

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