Context sensing and feature discovery for improving classifications

We propose context sensing as features for improved accuracy in classifications in our ongoing research. In many applications, features extracted from purposed sensors may not be enough for classification tasks accurately. Context features can help to discriminate classes in such cases. To address the problem, we first present how context can be used as features in classifications. Further, we present a case study on energy appliance identification from aggregated power usage for a household using context sensing and features.