Complex curved thin-walled structures, mainly using multi-axis milling, are highly susceptible to deformation induced by residual stress. It is therefore that there is a considerable amount of researches on developing predictive models for machining-induced residual stress. However, these developed models for residual stress prediction mainly focus on turning and threeaxis milling. For multi-axis milling, a hybrid model combining experimental results and finite element(FE) model is established to predict the residual stress profile of Ti-6Al-4V titanium alloy in the current study. Based on the experimental and simulated results, the residual stress profile is fitted by the hyperbolic tangent function using the firefly algorithm (FA). Good fitting accuracy is obtained, which the R 2 values change from 85.3% to 99.1% in the σx direction and change from 80.7% to 98.1% in the σy direction. The radial basis function (RBF) neural network is used to establish the relationship between the coefficients of hyperbolic tangent model and the milling parameters. The prediction accuracy is verified to achieve 92.7% and 91.4% in the σx and σy directions, respectively. The effects of cutting speed, feed rate, and inclination angle on surface residual stress and influence depth are investigated. The results show that there is a strong nonlinear relationship between the surface residual stress and milling parameters. The proposed hybrid prediction model of residual stress can be used for further machining optimization of complex curved thin-walled structures.