Research And Application Of Support Vector Machines In The Technical And Economical Evaluation Of Ship

Along with the development of ship industry and computer technology, lots of research on adopting prediction methods,optimization methods and multi-criteria evaluation methods to predict the future transportation quantity,to solve the main dimensions of ships and to carry through the ship-type evaluation have been done and made outstanding achievements. But the current methods are quite limited, and the result of technical and economical evaluation of ship depends on the chosen method too much, so some farther research is needed to do.The Support Vector Machines (SVM) brought forward by Vapnik and his research group is a new general machine learning method based on the VC dimension theory and the Structural Risk Minimization Principle of the Statistical Learning Theory, what it pursues is the optimal solution in the situation of existent information (the finite samples). This method is divided into two parts, SVC (Support Vector Machines for Classification) and SVR (Support Vector Regression), and it can solve the practical problems such as limited samples,high dimension,non-linear problem and local minimum. Because of their solid theoretical background and excellent generalization performance, they have become the hotspot of machine learning. Recently, Support Vector Regression (SVR) approaches the non-linear function with the controllable precision, and has the global optimization and excellent generalization performance. But the research in application is quite limited. Due to its outstanding learning performance and latent application value, the Support Vector Machines is attempted to apply into the fields of technical and economical evaluation of ship in this dissertation.The main contents and achievements of this dissertation are as follows:1. Based on the survey, the application actuality of the freight prediction,mathematical modeling of ship's principal dimensions and ship-type evaluation in the field of technical and economical evaluation of ship have been analyzed. Also the developments and actuality of the Statistical Learning Theory and the Support Vector Machines have been analyzed.2. The theory basics of Support Vector Machines have been detailedly introduced which contain the optimization theory, Statistical Learning Theory and kernel function theory.3. The elaborate deducing processes of the Support Vector Machines for Classification (SVC) and Support Vector Regression (SVR) have been introduced in detail in this dissertation. Some other existing forms of SVC and SVR have been simply introduced. The training algorithms and the characters of Support Vector Machines have also been summarized.4. A new algorithm has been proposed in this dissertation named Single-parameter Lagrangian Support Vector Regression and the elaborate deducing processes of this new algorithm are brought forward together. This new algorithm has also been applied to the artificial data set and practical data set for verification.5. Another new algorithm has been presented in this dissertation named Gaussian kernel parameter weighted Support Vector Regression. The artificial data set and practical data set have also been used to verify the efficiency of this new algorithm.6. The improved algorithms of Support Vector Regression proposed in this dissertation have been introduced into the field of freight prediction. Dividing into Time-series prediction and influence factor prediction, practical examples have been applied to prove the applicability and superiority of the algorithms.7. It is the first time bringing the Support Vector Regression in the field of mathematical modeling of ship's principal dimensions. Comparing the ordinary modeling methods, the SVR has been used to set up the regression models of the length,breadth,draft,depth and light weight of ship.8. It is the first time introducing The Kernel Principal Component Analysis (KPCA) into the field of ship-type evaluation. And sample analysis validates the applicability of KPCA in this field. The results show that the application of Support Vector Machines in technical and economical evaluation of ship is not only feasible, but also has the smaller samples,higher precision and better non-linear data processing and modeling performance than other methods. And it is a novel and encouraging research direction.