Accelerating Convergence in Data Offloading Solutions: A Greedy-Assisted Genetic Algorithm Approach

Data offloading, a technique that distributes data across the network, is crucial for alleviating congestion and enhancing system performance. One challenge in this process is optimizing web caching, which can be modeled as a dynamic knapsack problem in edge networks. This study introduces a Greedy-Assisted Genetic Algorithm (GA-Greedy) to tackle this challenge, accelerating convergence and improving solution quality. The greedy heuristic is integrated into the GA at two stages: during initialization to create a superior starting population, and at the end of each iteration to refine solutions generated through genetic operations. The GA-Greedy’s effectiveness was evaluated using the IRcache dataset, focusing on hit ratio—an indicator of successful cache accesses that reduces network load and speeds up data retrieval. Results show that GA-Greedy outperforms traditional GA and standalone greedy algorithms, especially with smaller cache sizes. For instance, with a 3K cache size, the half-greedy GA achieved a hit ratio of 0.55, compared to 0.2 for the pure GA and 0.1 for the greedy algorithm. Similarly, the full-greedy GA reached a hit ratio of 0.45. By enhancing convergence and guiding the search, GA-Greedy enables more efficient data distribution in edge networks, reducing latency and improving user experience.