Based On Incremental Learning, Support Vector Machine Classification Algorithm Research And Applications

This thesis studies the incremental training algorithm of support vector machine(SVM)and its application in the classification of sedimentary facies in the oil field.Based on the statistical learning theory and optimization method, support vector machine has becomes a new tool to solve the problems in machine learning area. It overcomes some shortcomings of neural network effectively, such as slow convergence, unstable solution, and bad generalization. Recently, SVM has become a hotspot of research in the area of machine learning.With the advancement of technology, the scale of information and data that to be processed is larger and larger, the same time, it is very difficult to get a complete data set at the beginning of training. Therefore, the incremental support vector machine classification technology is arousing more and more attention. For the above situation, the researches included in the thesis can be summarized as follows:Based on the thorough analysis of support vector machine algorithm, an improved support vector machine incremental algorithm is studied to solve the problem in the classical support vector machine incremental algorithm Batch SVM. It makes use of the condition of KKT to eliminate the samples in the initial set, and finally improves the classifying accuracy. Experiments show the validity of this method.The problem of pre-extracting training set is studied to decrease the training time of large scale support vector machine. Firstly of all, taking into account the geometric visual interpretation of support vector machine, a pre-extracting effective samples method based on the hyperplane nearest rules is presented. And than, on the basis of this method, combined with the first half part of the incremental improvement algorithm, a new support vector machine incremental method based on pre-extracting effective samples is presented. The computational complexity of the algorithm is analyzed theoretically. At last, the analysis of the algorithm s performance is tested by experiments.The application of incremental training algorithm of SVM based on pre-extracting effective samples in the classification of sedimentary facies in the oil field is discussed. The pretreatment and feature selection and extraction in the process of classification are described in detail. Experiments with log data show the validity of the algorithm.At last, summary of this paper is made and the problems needed further research are pointed out.