This article proposes an efficient strategy based on machine learning to address the challenges of spectrum resource management and allocation in 5G networks. We have implemented collaborative spectrum sensing using machine learning algorithms such as vector machines, decision trees, and logistic regression to improve the efficiency of spectrum resource utilization. And creatively proposed integrated algorithms, combining these three algorithms with voting algorithms to further optimize the accuracy and reliability of spectrum sensing. For the intelligent allocation problem of non occupied spectrum, the random forest algorithm is introduced to achieve intelligent allocation of spectrum resources, effectively solving the problem of low spectrum resource utilization. The comprehensive application of these methods provides innovative solutions for spectrum management and resource allocation in 5G networks. However, there may be some challenges in the implementation process, such as optimizing algorithm performance and system stability, which require further research and practice to solve. This article provides a promising research direction for spectrum management and resource allocation in 5G networks, as well as valuable experience and inspiration for future exploration in related fields.