DEAL: Dynamic Ensemble Algorithm for Imbalanced Streaming Data in UAV-Assisted Consumer IoT

Integrating the Internet of Things (IoT) in the consumer electronics market has led to the widespread deployment of connected smart devices, which generate a large amount of real-time sensing and monitoring data. In UAV-assisted consumer electronics, the dynamic environmental data collected by IoT-enabled drones changes over time due to changing conditions such as weather, obstacles, and user interactions, resulting in data imbalance and concept drift problems, affecting the accuracy of artificial intelligence models used for drone navigation, environmental monitoring, and other consumer applications. Effectively handling these constantly changing imbalanced data streams is critical to ensure reliable decision-making and adaptive drone operations. Therefore, we propose a dynamic integrated data stream method to solve the classification problem in the consumer IoT. First, we propose a mixed sampling method to balance the proportion of majority and minority class samples in the data stream and solve the sample imbalance problem in the consumer IoT data stream. Then, we use the chunk-based method to train the base classifier and merge it with the historical base classifier to form a candidate classifier pool, which will overcome the concept drift problem in the data stream and improve the performance of data stream classification. We conduct comparative experiments on IoT-driven UAV surveillance, smart home monitoring data streams, and synthetic benchmark datasets. The results show that our proposed method improves the F1 score by 2% compared to the data stream classification methods.