DAC-SNN: A Neuromorphic Adaptive Spiking Neural Network for Predictive Quality Assessment in Smart Campus Systems

Smart Campus quality assessment requires predictive models that can capture nonlinear interactions and temporal fluctuations across heterogeneous institutional indicators.Traditional machine-learning models operate on continuous activations, while conventional Spiking Neural Networks (SNNs) lack adaptive mechanisms to handle asynchronous spike dynamics.This study introduces the Dynamic Auto-Calibrating Spiking Neural Network (DAC-SNN), which integrates four mechanisms: Dynamic Threshold Adaptation, Hybrid Spike Encoding, Spike Synchronization Calibration, and Adaptive Hebbian-STDP Learning.All experiments were conducted under a unified evaluation protocol using Repeated Stratified 10-Fold Cross-Validation (10×10 CV) with leakage-free preprocessing.DAC-SNN achieved competitive regression performance (R² = 0.865 ± 0.017, MSE = 0.448 ± 0.022, MAE = 0.503 ± 0.021), outperforming Random Forest, Decision Tree, and MLP, while performing closely to the strongest baseline (SVR: R² = 0.881 ± 0.014).For classification, DAC-SNN reached 0.972 ± 0.011 accuracy.Ablation and sensitivity analyses confirm that spike synchronization provides the largest performance gain.