Data mining and knowledge discovery from physiological sensors

We present in this paper our method for discovering logic rules from physiological data by applying a fusion of preprocessing and data mining using decision trees. We have focused on four sensors representative of wearable technology capabilities namely; plethysmography, galvanic skin response, respiration, and body temperature, sourced from a highly cited dataset in Affective Computing. The method involved generating a number of datasets from the physiological data and subsequently performing classification using the C4.5 decision tree algorithm with a focus on knowledge discovery. The findings of this research demonstrate that preprocessing data into three classes with extreme boundaries and a neutral class, as well as classifying the two classes without the neutral class can produce high-quality rules. The discovered knowledge in the form of the top 4 rules was mapped on the valence and arousal emotional model. Finally, these rules are interpreted with the aid of box and whisker plots in the context of the underlying physiological processes.