Quantitative description of the complex relationship between rolling process and microstructure is crucial for improving the performance of 2205 duplex stainless steel products. In this work, four distinct rolling processes were used for three passes of hot rolling of 2205 duplex stainless steel. The study delved into the macroscopic morphology, microstructural gradients, and texture evolution of 2205 duplex stainless steel sheets in various rolling states. In addition, a data expansion method was used to expand the original experimental data, and the support vector machine model optimized by genetic algorithm was used to accurately describe the complex relationship between different rolling processes and microstructures. The results indicate that the proposed variable thickness cross-rolling method renders a more uniform microstructure in the rolled piece, with significantly refined grains and the formation of a distinct structural gradient. Among them, when using process 3 rolling, the grain size in the edge area of the plate after three passes of rolling is finer and the degree of recrystallization is higher, which is conducive to improving the problem of poor thermal plasticity at the edges and providing a basis for studying the suppression of edge crack damage in 2205 duplex stainless steel. In addition, the proposed machine learning model accurately predicted the grain size of plates in different rolling states, aligning closely with experimental findings.