In this work we propose a generalization of the Spherical Support Vector Machine method, in which the separator is a sphere, applied to Interval-valued data. This type of data belongs to a more general class, known as Symbolic Data, for which features are described by sets, intervals or histograms instead of classic arrays. This paradigm is raising interest in our days, specially in the context of Big Data. On the other end, SVM is a classic and well studied method for classification, and in the classical approach the separation is defined by an hyperplane. Generalizations of this concept have merged, with the use of non-linearity defined by Kernel functions. As an alternative to Kernell functions some authors have proposed non linear separation functions, in particular spherical separations, with the advantage of keeping the classification in the feature space. In this paper we present a quadratic optimization model for the spherical SVM for interval data and develop a relaxation linear problem. The performance of these two formulations for classification purposes is tested against standard methods for a collection of data sets.