ABSTRACT
Consider the seamless and efficient operation of potent AI applications like image recognition or natural language processing on your mobile device. The system leverages pre-built or custom AI models, often demanding significant processing power. It dynamically assesses your device's capabilities, including battery level, CPU usage, memory consumption, and network strength. Based on this real-time analysis, it decides whether to execute the task locally or offload it to robust infrastructure. This ensures smooth operation and optimal resource use, mitigating concerns about battery drain or intermittent performance. The system integrates a machine learning model that monitors device parameters, enabling smarter offloading decisions. This dynamic approach ensures peak performance, responsible resource management, and a seamless user experience. The study utilised algorithms such as K-Nearest Neighbours (KNN), Ensemble Stacking, Random Forest, and Gradient Boosting to optimise the offloading process. An analysis of key performance metrics, including accuracy, specificity, sensitivity, F1-score, and precision, along with execution time, underscored the effectiveness of gradient boosting. The gradient boosting algorithm achieved an accuracy rate of 97.6%, with an average execution time of 3 ms, and performed competitively in all other metrics, demonstrating its efficiency in optimising the offloading process for mobile AI applications.
