Application of Support Vector Machines to a Small-Sample Prediction

摘要

The support vector machines (SVMs) is one kind of novel small-sample machine learning methods based on solid theoretical background. Highly nonlinear regression and classification are their two applications. Different from conventional statistics methods, the SVMs employs the structural risk minimizing principle, which leads to high predication precision. For this method is not essentially related to probability measure and Law of Large Numbers, the final decision function is only determined by a small fraction of sample, called support vectors. Consequently, the complexity of computation only depends on the number of support vectors rather than the dimensions of the original sample space. In most occasions of oil and gas development, only small samples are available to predict the results of one measure. Introduction of SVMs into these applications can significantly improve prediction precision.