University Archives

This paper presents an environmental monitoring platform using a Raspberry Pi 4 with dual sensors to capture temperature, humidity, atmospheric pressure, and CO₂ at 60-second intervals. The system delegates 85% of processing to the edge device, including real-time acquisition, validation, local storage, and alerting, while external devices support model training and visualization. The core contribution is an Isolation Forest model trained on 2–3 weeks of unlabeled sensor data, enabling anomaly detection without cloud connectivity or labeled examples. Rule-based thresholds complement the model for critical conditions such as extreme CO2, temperature bounds, and high humidity. A preprocessing pipeline enforces physically plausible ranges, applies smoothing, and normalizes timestamps. Results demonstrate stable sensor integration, low-latency communication, and effective real-time visualization. The platform illustrates how lightweight edge AI can deliver privacy-preserving environmental intelligence for smart homes, with applications to indoor air quality management and HVAC optimization.