A feature-based classification technique is applied to the analysis of solar wind properties upstream of the Earth to predict the occurrence, duration, and magnitude of magnetic structures that can cause large geomagnetic storms. Because this method is based on identifiable physical features, it is highly upgradable by either analyzing data or physical models. The formulation of the technique is discussed and is then applied to two relatively ''simple'' solar wind events leading to large storms. The introduction of additional features to classify more ''complicated'' events is discussed to illustrate the upgradability.