To investigate the frost damage of Zoige Plateau Peat Permafrost and the stability of engineering foundations in seasonally frozen regions, this study builds upon previous research on the optimal mix ratio of composite stabilized peat soil using mineral powder-steel slag and basalt fibre as primary materials, supplemented with polycarboxylate superplasticizer and carbide slag. Indoor mechanical tests (unconsolidated undrained triaxial shear test, unconfined compressive strength test, and scanning electron microscopy test) were conducted on stabilized soils subjected to varying freezing temperatures and freeze-thaw cycles to explore the mechanical properties and microscopic mechanisms under freeze-thaw action. Based on the experimental dataset, the peak strength of stabilized peat soil was predicted using a particle swarm optimization backpropagation neural network (PSO-BP). The results indicate that the unconfined compressive strength and residual strength ratio of stabilized peat soil are positively correlated with freezing temperature and negatively correlated with the number of freeze-thaw cycles. The stress-strain curve characteristics of stabilized peat soil are generally consistent under different conditions. With increasing cycles, the peak strength and cohesion of the stabilized peat soil first decrease and then stabilize, while the internal friction angle remains relatively unchanged. In the peak strength model constructed using MATLAB, the PSO-BP neural network model exhibited a higher correlation coefficient (R²) and better overfitting performance compared to the BP neural network model, enabling a more accurate strength prediction of stabilized peat soil.