Environmental Hazards Prediction using Real-Time IoT Sensor Data and Edge-Deployed Machine Learning for Smart Residential Contexts

Abstract

The rapid urbanization and increasing environmental concerns necessitate innovative solutions for smart home security that extend beyond traditional humancentric threats. This paper presents an Internet of things (IoT) based smart home security system designed to monitor and mitigate environmental hazards such as gas leaks, extreme temperatures, humidity fluctuations, and fire risks. The system integrates multiple sensors (gas, temperature, humidity, motion) and a pan-tilt camera module connected to a Firebase cloud backend for real-time data processing and storage. A dedicated mobile application enables users to access live sensor readings, camera feeds, historical data visualization, and instant hazard alerts. Machine learning algorithms implemented through Random Forest and Isolation Forest models analyze sensor patterns to detect anomalies and predict hazardous conditions. Experimental evaluation over a 30-day period across 15 households achieved 91.2% hazard classification accuracy, 98.7% data transmission reliability, and an average response latency of 320 miliseconds, while reducing cloud data load by 62% through edge inference. The system demonstrates robust real-time monitoring, early hazard prediction, and user satisfaction ratings averaging 4.3/5 for usability. This research contributes to advancing smart home technologies by addressing critical environmental security gaps through an integrated, data-driven, and edge-enabled IoT framework.