Hybrid Performance Prediction Model for Edge Devices

As the adoption of edge devices for computationally intensive applications grows, the need for accurate and efficient performance prediction models becomes critical. This paper presents a hybrid performance prediction model designed to measure the performance of applications running on edge devices by combining static and dynamic (runtime) analysis. In contrast to models that utilize static or dynamic data, the hybrid model we present in this work estimates the performance of edge devices and proves to be more precise compared to only a static analysis-based approach. The proposed model also addresses the limitations of relying solely on a dynamic model. MAPE of the proposed model is 19.75% and the RMSE is 1.912 ms.