A key issue in the practical applicability of the support vector machine methodology is the identification of the support vectors in very large data sets, a problem to which a great deal of attention has been given in the literature. In the present article we propose methods based on sampling and nearest neighbors, that allow for an efficient implementation of an approximate solution to the classification problem and, at least in some problems, will help in identifying a significant fraction of the support vectors in large data sets at low cost. The performance of the proposed method is evaluated in different examples and some of its theoretical properties are discussed.