PSO-SVM Learning Algorithm And Its Application In The Analysis Of Spatial Data

With the rapid development of science and technology, humanity has entered thedigital era. Today, how to find and analyze the laws from massive spatial data hasbecome a hot research, which is of great significance on the human study anddevelopment. The traditional methods, such as statistical methods, artificial neuralnetwork algorithm, etc., want to obtain better results which need to have enough thenumber of data samples. The number of collected samples is often limited in practicalproblems, so the traditional methods are difficult to achieve the desired results. However,support vector machines algorithm is a new data-based machine learning algorithms,which designed specifically for small samples. This algorithm is based on theVC-dimensional theory and structural risk minimization theory on statistical learningtheory. On solving small sample, nonlinear and high dimensional pattern recognitionproblem, support vector machines algorithm shows a lot of unique advantages, so thismethod has been widely used in a various kinds of fields, such as pattern recognition,regression estimation, probability density function estimation, character recognition,speech recognition, text categorization, signal processing, etc., however, support vectormachines algorithm also has some shortcomings, for example how to select the functionsand parameters of support vector machines has a direct impact on the performance of thealgorithm. In view of the above problems, the support vector machine algorithm based onhybrid kernel and PSO is proposed. Then this improved algorithm is respectively appliedto the typical binary classification problem and multi-classification problem. AfterMATLAB simulation experiment, the achieved results are satisfactory. The paper hopesto provide a new idea for solving classification problem of spatial data.In the first chapter, the paper analyzes the subject background and the significanceof the topics, finally, the main content and structure of the paper are arranged.In the second chapter, in order to better understanding the basic principles ofsupport vector machine and improved algorithm, the paper introduces the basictheoretical knowledge of the support vector machine algorithm from the optimizationtheory and statistical learning theory. In the third chapter, considering the typical binary classification problem andmulti-classification problem, the paper analyzes the basic principle of the support vectormachine algorithm, and then, the support vector machine algorithm based on hybridkernel and PSO is proposed, because the support vector machine algorithm has twoaspects of disadvantages, which are the choice of parameters and kernel function.In the last chapter, considering the binary classification problem and themulti-classification problem, the support vector machine algorithm based on hybridkernel and PSO is applied to cancer diagnosis on the UCI breast_cancer database andface recognition on ORL face database, the former problem is a typical binaryclassification problem, the last one is multi-classification problem. Finally, afterMATLAB simulation experiment, the results show that the improved algorithm hashigher classification accuracy and better learning and generalization ability.