Implementation Of Transductive Support Vector Machine In Data Prediction

Support Vector Machine is studied as much as possible in recent years. In our thesis, SVM application is enlarged into three fields which include velocity prediction by SVM regression, texture synthesis by TSVM and object recognition by TSVM. These three fields'essence is data prediction by SVM, which shows its promising value in the future.For velocity prediction, SVM regression is firstly used in oceanography . Ocean is a treasure house for human beings, which provide us a mount of aquatic product. Sea water exchange is an important source of nutrient salts. In recent years, statistical learning theories represented by SVM Regression have been well developed. However, no publications are available regarding using SVM regression to predict marine environment elements related to hydrodynamic and combining these predicted elements with ocean models. So, using SVM regression to predict velocity of wave which resisted by seaweed has a profound meaning.In texture syntheses field, TSVM is firstly applied in self-similar synthesis. Texture synthesis has broad application areas and meanings. TSVM uses labeled and unlabeled feature vectors to learn a model. The unlabeled feature vectors would be assigned the proper label during retraining. Using less time makes TSVM in texture synthesis has a more promising future.For object recognition , it is widely used in intelligent security control and traffic monitoring. TSVM is applied with tree strategy using HOG and Gabor feature. After a lot of experiments, the results are compared with the results of active basis method. The new method could obtain a better result.