Utilizing Stacked Ensemble AI Algorithm for Comprehensive Water Quality Analysis and Health Assessment in Complex Aquatic Systems

Abstract

Abstract A novel framework, referred to as the Intelligent Mobile Data Collection (MDC) framework, is proposed to enhance data collection efficiency in Internet of Things (IoT) based sensor networks. This framework organizes IoT devices and sensors into clusters based on their geographical proximity or region. Within each cluster, a gateway node is designated to collect and consolidate data from its constituent members before transmitting it to the central MDC. To optimize data collection, the framework employs a learning mechanism known as Frequency-Based Reinforcement Learning (FRL). This technique analyzes data generation patterns, such as time intervals between transmissions, the quantity and type of packets generated, to classify clusters into categories: Frequent, Less Frequent, Rare, and Very Rare. Within FRL, each IoT sensor or device independently trains its local model using Reinforcement Learning (RL) techniques, encompassing states, actions, and rewards. These local models capture the specific behaviors and characteristics of the sensors. Subsequently, IoT sensors transmit their local model parameters to the gateway, where they are aggregated into a global model. This aggregated global model is then disseminated back to the IoT sensors, enabling them to adjust their behavior based on collective insights. Based on the categorized clusters, the framework dynamically adjusts parameters such as Time Division Multiple Access (TDMA) slot allocations, sleep durations for sensors, and the visiting schedule of the MDC. This adaptive approach ensures efficient utilization of network resources while accommodating varying data generation rates and priorities across different clusters. In summary, the proposed Intelligent Mobile Data Collection framework integrates FRL and RL techniques to optimize data collection in IoT sensor networks. By dynamically adapting to changing data generation patterns and cluster characteristics, it enhances overall network performance and resource utilization.