Machine Learning Techniques for Context-Aware Data Aggregation and Filtering in the Internet of Things

Abstract

The rapid growth of IoT devices has generated huge volumes of heterogeneous data, which poses major challenges in terms of data aggregation, filtering, network performance and real-time decision making. This paper presents a framework based on machine learning for context-aware data aggregation and filtering in the IoT environment using supervised, unsupervised and reinforcement learning techniques. The proposed framework integrates context extraction, pre-processing, adaptive filtering and intelligent data aggregation to enhance the effectiveness of the data management of the Internet of Things. The experimental evaluation was performed with the NSL-KDD dataset and implementations were developed with the Python U-learn and TensorFlow libraries. Five machine learning techniques, namely reinforcement learning (RL), decision tree, support vector machine (SVM), K-means clustering, and artificial neural network (ANN), were assessed for their data reduction rate, aggregation efficiency, throughput, accuracy of context detection, latency, scalability, and fault tolerance. The experimental results show that the ANN achieves the highest aggregation efficiency (89%), filtering accuracy (95%), context-detection accuracy (91), energy efficiency (51) and processing efficiency in real time (11) while maintaining the lowest latency (23ms) and data processing overhead. Reinforcement learning demonstrated greater adaptability to the dynamic IoT environment and achieved the highest error tolerance (91%), which makes it suitable for context-aware adaptive applications. The findings show that ANN is well suited for the computationally intensive applications of the Internet of Things, while RL offers better adaptability to a dynamic and resource constrained environment. The proposed framework improves intelligent data management, reduces redundant transmission, improves the use of resources and supports the development of scalable, secure and context-aware systems for the Internet of Things.