Edge-Driven Intelligence: A Hybrid Machine Learning Strategy for IoT Edge Nodes

The efficient deployment of machine learning (ML) inference models on the Internet-of-Things (IoT) edge nodes faces significant challenges due to hardware limitations. The limited computational resources, including computational power and memory, restrict the direct implementation of complex ML algorithms on these devices. Hence, there is a need to explore optimized, and resource-efficient approaches that can leverage the potential of ML while accommodating the constraints of IoT edge hardware and ensuring intelligent decision-making at the IoT edge. We propose a hybrid inference model that incorporates Kernel Principal Component Analysis (PCA), Extreme Gradient Boosting (XGBoost), and Support Vector Machine (SVM) to address this issue. By integrating Kernel PCA, we perform efficient feature extraction and dimensionality reduction through its ability to handle nonlinearly separable data by utilizing kernel functions. This enables the model to handle high-dimensional data effectively. The XGBoost algorithm further enhances the model’s predictive capabilities by offering superior gradient boosting and ensemble learning to handle complex and large scale datasets. Additionally, SVM is incorporated to harness its strong generalization capabilities for accurate and robust decision-making. The proposed hybrid inference model optimally balances computational efficiency and accuracy, making it well-suited for real-time deployment on IoT edge nodes. Through extensive experiments and evaluations, we demonstrate the efficacy of our approach, highlighting its potential to significantly advance IoT edge computing.