Classification Of Genes Based On Least Squares Fuzzy Support Vector Machines

With the rapid development of molecular biological science and information science, bio-information technology becomes the forefront as a new area of research. A lot of gene expression data can be divided into a relatively small number of groups with biological meaning using genetic classification, and on this basis value from the information extracted it can clarify the functions of different genetic composition and working methods. How to carry out an effective gene classification of data to extract on the physiology and medicine and other valuable information for biological has been a hot research spots. Therefore, this thesis, the fuzzy support vector machine classification of gene technology has practical significance.Fuzzy support vector machine which has advantages of SVM generalization ability with the global optimum and the advantages of fuzzy technology, such as non-relying on fuzzy plant model and strong point of robustness, has been widely developed in recent years. In this thesis, the fuzzy support vector machine which has been applied to gene classification, can be effectively distinguish the samples of noise from the samples of outliers. This method improves the accuracy of the classification and has a good feasibility.In this thesis, the focus of the study is gene classification and the algorithm for fuzzy support vector machine. Here, we propose the algorithm of fuzzy support vector machine and apply it to gene classification. Attention on the samples near to classification surface, we propose classification algorithm based on fuzzy least squares support vector machine. In the classification process, the sample points near to the separating hyper plane might become the support vector and affect the results of the classification. Therefore, we should set larger degrees in the design of membership, and the experimental results show that this algorithm improves the accuracy of classification.Fuzzy least squares support vector machine has quickly solve the QP question, in the process of design fuzzy membership, in addition to premeditate the relationship between the samples and the class center, the relations of the samples should also be taken into account. Thence, a density-based algorithm for least squares support vector machines has been proposed in this thesis. We determine the density through the number of samples in a certain range, then design of membership function, which can well reduce the effect of the noise and outliers on the classification of samples. And apply the algorithm to gene classification, experiments show that the algorithm can be effectively distinguished noise and outlier samples,and improve the classification accuracy.