Artificial intelligence, or edge AI, is being deployed at the network edge as an outcome of the growth of distributed computing and the expanding availability of mobile data. The combination of machine learning and wireless communication creates new difficulties for effective radio resource management. While all data points are handled equally in traditional communication, some data points have a bigger influence on the learning process in machine learning. It becomes crucial to devote more resources to the data that has a major impact on machine learning performance and use fewer resources for less important data in order to maximize the use of the limited radio resources. Human centralized edge machine learning, decentralized edge machine learning, there are specific criteria to assess the significance of data according to its influence on machine learning results. We present resource allocation techniques designed to enhance edge AI performance based on these standards. These information-essential alert radio assets assignment plan to have been proven to be effective through extensive experiments, showing promise for improving machine learning tasks in wireless networks. Two criteria for evaluating the significance of data based on its effect on learning accuracy are identified in the proposed work, single for distributed machine learning and one for centralized edge machine learning. By using these importance metrics, we create radio resource management plans that give high-impact data transmission top priority, improving performance for edge AI applications. Better performance was shown by the simulation results in terms of connection overhead, latency, bandwidth consumption, and network lifetime.