Dissertation > Excellent graduate degree dissertation topics show

Conformal Predictor Based on Transduction

Author: WangLi
Tutor: DingXiangQian
School: Ocean University of China
Course: Signal and Information Processing
Keywords: Conformal Predictor Transduction Algorithm Randomness Theory Nonconformity Measure Geodesic Distance Logistic Regression
CLC: TP181
Type: Master's thesis
Year: 2011
Downloads: 18
Quote: 0
Read: Download Dissertation

Abstract


Some classification algorithm in machine learning are based on induction and deduction, when facing to a batch of data, they are always trying to extract a general classification model for those data, and expect the classification model has higher prediction performance; But in high-risk and small sample fields, this classification model have some limitations:firstly, because of the limitation of the number of small sample data, this classification model not only didn’t have enough data used for extract a general rule, but also can not make full use of all the information of sample data; Secondly the results of classification prediction are just simple to us, and the results have no confidence or believability to measure.In this paper the research and application of conformal predictor based on transduction is aimed at this defect to launch. Transductive inference is directly to classify and recognize for the data sets with the whole sample information, and it doesn’t need to build classification model in the process of classification, its target is to explore the inner relation between tested sample data and sample sequence; The conformal predictor build a confidence mechanism through the Kolmogorov algorithm randomness theory, and use the detect function of randomness whose results are the reliability of prediction, to measure the confidence, and the form of results predicted by algorithm under the control of confidence are domain.The mainly part of conformal predictor is to design the function of nonconformity measure, which can be integrated classification algorithm in traditional machine learning. In this paper, firstly the conformal prediction algorithm based on k-nearest neighbor and support vector machine are discussed, and the conformal prediction algorithm based on k-nearest neighbor is improved, through using geodesic distance to replace Euclidean distance to measure the similarity between samples; Then the conformal prediction algorithm based on logistic regression model can be achieved; All algorithms in this paper are programmed on the platform of MatlabR2007b, and be used to the data sets from UCI and tobacco industry, and through the comparison of accuracy rate predicted by conformal prediction algorithm and other algorithm used usually to test the performance of conformal prediction algorithm.In the end, the thesis summarizes the research and indicates the future work.

Related Dissertations

  1. 40-59 years old coastal city of Shandong Province overweight, obese population distribution , physical characteristics and influencing factors,G804.49
  2. The Logistic Regression Analysis of High Frequency and Color Doppler Flow Imaging in Thyroid Solitary Nodules,R445.1
  3. Retail moral hazard perception impact study on consumer behavior,F274;F224
  4. The Research on Transduction Efficiency of Surface Wave Emat,TB552
  5. Comparative Research on Models of Financial Distress Warning in Chinese Listed Companies,F276.6;F224
  6. Based on CT images of peripheral non -small cell lung cancer diagnosis of type Model,TP391.41
  7. Acute Exercise Induce PGC-1α Gene Transcription in Skeletal Mucle of Rat,G804.2
  8. Types of casual menswear customer segmentation,J523.5
  9. Study on the Transduction of Gm6 Gene Rice Gall Midge Resistance by Using the Sanming Dominant Genic Male Sterile as the Vector,S435.112
  10. Application of GEE and MLM to Case-crossover Study,R195
  11. Analysis on Hearing Loss of Workers under the Noise Environment in a Steel Corporation and Its Influencing Factors,R131
  12. Hermap erythroid differentiation genes during the development of cellular signal transduction pathway in research,R555
  13. Silicosis and pulmonary fibrosis in rats during the critical difference in gene expression and signal transduction pathways Discussion,R563.9
  14. Formation of aldosterone in the metabolic syndrome,R589
  15. The Role of p38MAPK on High-glucose Induced ICAM-1 and VCAM-1 Expression in Mesangial Cell,R587.2
  16. Rice plant sulfur peptide hormone gene transfer system to build and Traits,S511
  17. Optimizing the Techno-system of Genetic Transformation Mediated by Agrobacterium Tumefaciens and LeJA2 Gene Function Research of Tomato,S641.2
  18. Introduction reject inference Small Business Credit Scoring Model,F276.3;F224
  19. The Research on the Dynamic Early-Warning of Credit Risk Based from the Commercial Banks’ Perspective,F830.33
  20. Analysis on Function of PP2CA2 Gene T-DNA Insertion Mutant in Arabidopsis,Q943.2
  21. Peasant Workers Training Wishes and the Affecting Factors Analysis Based on the Investigation of Fuzhou,F224

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