After the optical micrography (OM) and scanning electron microscopy (SEM) observations, the grain size of primary α phase was measured via a quantitative metallography image analysis software. The effect of deformation temperature and strain rate on the microstructure was discussed. A Pi-sigma fuzzy neural network (FNN), in which the layers of neural networks were organized into a feed-forward system, was used to predict the flow stress and the grain size during isothermal compression of Ti-6Al-2Zr-2Sn-2Mo-1.5Cr-2Nb alloy. The comparisons of the predicted flow stress and grain size for the sample data or the non-sample data with the experimental results were given to train the models and confirm the validity in present study. The results show that the accuracy of prediction from the Pi-sigma FNN models is much high, and the Pi-sigma FNN approach can efficiently describe the non-linear and complex relationship of titanium alloys.