The speedy rise of networked systems in the cloud has posed great challenges to resource allocation, latency control and power economy under dynamic workloads with diverse heterogeneities. Conventional static and rule-based resource management mechanisms do not always accommodate dynamic changes in network traffic and application requests on the fly. This work introduces an intelligent resource optimization methodology for cloud-based networked systems based on machine learning. The newly proposed model uses predictive analytics and adaptive learning algorithms to achieve real-time optimization of computing, storage, and network resources. Performance metrics in terms of resource utilization, response time, throughput, and energy efficiency are dynamically monitored and optimized by supervised learning and reinforcement learning models. Experimental results in simulated cloud-network environment show that the new approach greatly improves resource utilization, reduces latency, and lowers operation overhead with respect to state-of-the-art optimization methodologies. The presented findings validate the utility of machine learning based decision making to realize autonomous, resilient, and scalable cloud networks management. In conclusion, the studied demonstrates that intelligent resource optimization can play a significant role in enabling next-generation cloud services like edge computing, IoT and large-scale distributed applications.