Effects of Kalman Filtering on KNN-Based Classification in RFID-Enabled Smart-Lab Real-Time Energy Monitoring

Smart laboratories require effective methods for user authentication and real-time energy monitoring. This paper investigates the influence of Kalman filtering on K-Nearest Neighbors (KNN) performance for classifying energy consumption levels from RFID-based sensor data. We continuously recorded voltage, current, and power data in two distinct lab setups, labeling each dataset into low, medium, and high energy categories. Tests indicated moderate Kalman filtering improved signal stability and KNN accuracy. However, excessive filtering reduced accuracy due to loss of essential signal details. Our best Kalmanfilter configuration achieved a 95% accuracy, surpassing random forest and decision tree methods. These findings highlight the necessity of precise Kalman filter tuning to maintain data quality and classification performance in RFID-enabled smart lab energy monitoring.