Enabling Disaster Predictions on IoT Scenarios with Transfer Learning

Information and communication technology have changed the way humanity sees the world around it. Faced with the maturation of the Internet of Things (IoT) and the possibilities offered by artificial intelligence, we have a more remarkable ability to observe and estimate things and situations based on information collected in the environment. This work investigates the use of transfer of learning as an enabling tool for monitoring remote locations,aiming at disaster prevention. The main objective is to provide a viable solution for bandwidthconstrained scenarios for data communication. As a way to validate the proposed architecture, we present results obtained in a prototype built in our laboratories that allows us toevaluate the accuracy of machine learning models under different conditions of transmission of monitoring data. The results demonstrate the effectiveness of the transfer of learning technique as one of the pillars for the scalability of a disaster monitoring system