A Fog Cloud Integrated Framework for Data Transformation and Machine Learning Analytics in IoT Systems

This study analyzes the efficacy of three data processing architectures: fog computing, cloud computing, and a hybrid fog-cloud computing framework within an Internet of Things (IoT) context. We evaluate the performance of machine learning methods regarding computation time, communication overhead, precision, and efficiency on IoT-based sensor data. The proposed hardware utilizes a Raspberry Pi as a fog node for local data processing and a cloud infrastructure for executing computationally intensive operations. Each architecture processes both raw and processed real-world IoT sensor data. The transformation procedure at the fog node compresses data, reducing its size and minimizing overhead during transmission to the cloud. The findings demonstrate that the hybrid architecture offers the best balance of execution speed, communication efficiency, and accuracy when compared to cloud-only and fog-only methods. The hybrid system leverages cloud computing capabilities and minimizes transmission costs by collecting data locally.