An Adaptive AI-Driven Code Optimization Framework Using Machine Learning and Explainable AI for Scalable Software Systems

The increasing complexity of embedded computing systems has made it difficult to achieve efficient code execution while effectively utilizing limited hardware resources. Conventional optimization methods mainly rely on compiler-driven techniques and fixed transformation rules, which may not adapt well to different embedded architectures and varying workload conditions. This research introduces a hardware-aware AI-driven code optimization framework that integrates Machine Learning (ML) with Explainable Artificial Intelligence (XAI) to improve the efficiency and scalability of embedded software systems. The framework analyzes software characteristics together with hardware-related parameters, including processor usage, memory requirements, computational workload, and resource constraints, to identify performance limitations and select suitable optimization strategies. A machine learning model is employed to predict appropriate code transformation techniques based on application behaviour and underlying hardware conditions. The XAI module further analyzes the optimization process by identifying the contribution of individual features, providing clearer explanations for the selected optimization decisions. Experimental results show that the proposed framework reduces execution time by up to 30%, decreases memory consumption by 25%, and lowers CPU utilization by approximately 15% while maintaining performance as code complexity increases. These findings demonstrate the capability of the framework to support efficient and transparent optimization of embedded software under resource-constrained computing conditions.

@article{K2026,
  title     = {An Adaptive AI-Driven Code Optimization Framework Using Machine Learning and Explainable AI for Scalable Software Systems},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {17},
  number    = {9},
  year      = {2026},
  publisher = {The Science and Information Organization},
  author    = {Kavya B K and S.Annamalai and A Suresh Kumar},
  doi       = {10.14569/IJACSA.2026.0170945},
  url       = {https://doi.org/10.14569/IJACSA.2026.0170945}
}