General-Purpose Sensing for Smart Environments: The Smart Museum Use Case

A comprehensive monitoring of the indoor environment opens the door to multiple and orthogonal application scenarios. In the context of a Smart Museum, for example, General-Purpose Sensing approaches with measurements like air quality, temperature, humidity allow estimating both usercentric metrics like visitors' comfort but, also, artifact-centric ones such as artwork preservation. Hence, in this paper, we present a CNN-LSTM (Convolutional Neural Network, Long Short-Term Memory) predictive approach to forecast indoor temperature variations within Smart Museum, enabling proactive microclimate management and the prevention of potential damage to artifacts. The results show that the predictive model achieves good accuracy in temperature forecasting, with a Mean Absolute Percentage Error (MAPE) of 2 % in the calculation of the Predicted Risk of Damage (PRD). This data-driven approach offers the potential to prevent damage to artifacts through proactive management of environmental conditions while simultaneously ensuring direct artifact supervision and the full exploitation of already deployed hardware.