The interlinked nature of mobile users and terminals has been significantly impacted by Wireless Communication System (WCS). A hybrid Particle-Swarm-Optimized Artificial Neural Network (PSO-ANN) framework is presented in this paper for proactive downlink interference mitigation in multi cell Orthogonal Frequency Division Multiple Access (OFDMA) systems. Large-scale route loss, instantaneous Signal-to-Interference-Plus-Noise Ratio (SINR), user mobility vectors, and neighboring-cell power budgets are all included in the compact feature vector that is taken from live radio measurements and fed into a multilayer artificial neural network. PSO globally adjusts the network's initial weights and biases prior to traditional back-propagation refinements in order to avoid poor local minima and speed up convergence. Simulations are conducted and the proposed PSO-ANN outperforms standalone neural networks and traditional reuse schemes in terms of signal quality, network throughput, and interference suppression. The method is resilient across a broad range of user mobility and traffic volumes while maintaining inference delays that are consistent with ultra-reliable low-latency communication needs. According to the results, PSO-ANN is a scalable and effective option for managing radio resources in future wireless infrastructures while taking interference into account.