An SVM is a numerical classifier that in some ways is similar to the KNN algorithm, although the SVM is far more mathematically advanced. Rather than comparing a test point to the points closest to it, an SVM attempts to draw boundary lines between the classes of data points, creating regions where all points inside that region will be considered a member of that class.
Consider this image (from Wikipedia's article on SVMs). The two categories of data points are separated by a straight line. The line that separates the classes is chosen as the line of maximum margin, meaning this dividing line has the most room on either side of it, as compared to any other separating line you can draw:
The SVM, exactly as implemented ...
