A kernel method lets a support vector machine operate in a transformed feature space without explicitly constructing every added feature. This is useful when data that cannot be separated by a linear boundary in its original representation may become separable in a richer one. section 8.1.2
Temperature readings
Safe readings in the middle and unsafe readings at both extremes cannot be separated by a single threshold. Mapping temperature to temperature and temperature squared makes the classes linearly separable; a kernel method lets an SVM work in such a transformed space without explicitly creating every feature. section 8.1.2
Figure 8.3 Mapping each temperature reading into a two-dimensional feature space defined by temperature and temperature squared makes the classes linearly separable. The points are schematic rather than exact functional values. The dashed line is a hyperplane separating the safe observations, shown as circles, from the unsafe observations, shown as squares. It demonstrates possible separation, not the optimal maximum-margin boundary.

See how adding temperature squared maps the readings into a feature space where a straight boundary can separate the classes. section 8.1.2

