Machine Learning Application for Automatic Emergency Signal Activation

Abstract

The development of an innovative system that uses Machine Learning and IoT sensors to automatically activate emergency signals in critical situations, improving the speed and efficiency of the response. Using a Random Forest machine learning model, trained with data from temperature, gas, humidity, and flame sensors, the system achieved a 96.8% accuracy, with key metrics such as an AUC of 0.997 and an F1-score of 0.968. Integrated with an Arduino microcontroller, this system can autonomously activate alarms and lights, eliminating the need for human intervention in emergency situations. By detecting risks such as gas leaks, fires, or temperature spikes, the system responds almost instantly, which can be crucial for saving lives. This approach not only optimizes safety in vulnerable environments but also establishes a smarter and more efficient model for emergency management.