Practical Approach of Fast-Data Architecture Applied to Alert Generation in Emergency Evacuation Systems

This paper describes a proof of concept of a Fast-Data architecture to generate early response alerts on unforeseen events. For achieving that, in this work is presented the implementation of a fully integrated system capable to handle and process streaming data in order to generate an alert response for each generated event. The deployment stated are composed by a simulated wireless sensor network for generating environmental values, a centralized Kafka server for data segmentation and a machine learning model deployed in a Spark cluster for generating the emergency alerts. Also, a simulation was conducted assuming that a fire had affected the simulated scenario in order to determine and evaluate the system's behavior. Finally, the classification model is presented as an early system alternative based on real-time processing and can be used in different areas of occupational safety.