Adaptive Intercell Interference Management in Heterogeneous Networks Using Enhanced Feedforward Neural Network: A Drive-Test-Based Simulation Study

The densification of Long-Term Evolution (LTE) heterogeneous networks (HetNets) enhances spectral efficiency but exacerbates intercell interference (ICI), especially at cell edges. This study proposes an Enhanced Feedforward Neural Network (eFFNN) for adaptive interference mitigation using 12,847 drive-test samples from Port Harcourt. The model integrates advanced techniques such as batch normalization, dropout, and dynamic loss weighting, and is deployed within a MATLAB/Simulink-based LTE HetNet simulation. Results show significant performance gains, including 68% reduction in packet loss, 36.8% improvement in throughput (85.48 Mbps), and 26–32% latency reduction. The model achieves high predictive accuracy with an R² of 0.9969, demonstrating strong effectiveness.