Adaptive Anomaly Detection in Sensor Data: A Comprehensive Approach
Tanya Singh, Snigdha Nigam, Eshita Vijay, Ridhima Rathore, Snehal Bhosale, Abhijeet Deogirikar · 2023
In the expansive realm of the Internet of Things (IoT), ensuring the reliability of sensor data, particularly in level sensing applications, is paramount. This research introduces a pioneering solution-a generalized anomaly detection algorithm rooted in supervised machine learning principles. Our algorithm, diverging from traditional rule-based systems, leverages labeled training data to dynamically adapt to unique patterns across diverse sensors. Live data, with a focus on level sensors, forms the cornerstone of our research, enhancing the algorithm’s learning capabilities for precise anomaly identification in various operational environments. Practically implementing our algorithm involves the systematic storage and management of sensor data in a MySQL database. This dual-purpose database supports both the algorithm’s training and testing phases and serves as a historical repository for continuous refinement, ensuring ongoing adaptability to changing sensor behaviors. Simultaneously, we introduce a user-friendly dashboard, empowering end-users with actionable insights derived from anomaly detection. This visualization tool provides a comprehensive overview of sensor data, anomaly occurrences, and system health, facilitating rapid decision-making and proactive intervention. Our research pioneers a breakthrough in anomaly detection for level sensor data with a generalized algorithm grounded in supervised machine learning. By leveraging live data, a robust MySQL database, and an intuitive dashboard, our approach stands at the forefront of AI-based anomaly detection, offering adaptability across sensors and applications. This research not only advances level sensing but also establishes a foundation for a more efficient paradigm in anomaly detection across diverse sensors and IoT devices.