Abstract
With the evolution of sensor-empowered systems in fields ranging from smart cities (e.g., green energy) and industry (e.g., machines condition monitoring) to personal and social health and wellness (e.g., wearable devices and electroencephalography, EEG), it is vital to consider constraints related to these sensors, including power consumption and noise. Consequently, empowering intelligence (e.g. prediction of states or outcomes) in such systems becomes of high importance in order to better deal with such constraints and to reduce the effects of dimensionality and volume of sensed data. Here, we propose a sensor selection algorithm that minimizes the sensor space relying on the similarities in how sensors sense a given environment. This algorithm resulted in a significant reduction in the sensor space without notable loss of accuracy when compared against the complete set of the sensing space. We applied the proposed algorithm to two sensor-based time-series datasets of different fields of application.