In the process of milling titanium alloy, workpiece surface roughness is mainly affected by cutting parameters, tool angle, tool shape and system vibration. There is a complex highly nonlinear relationship among these factors, which makes it impossible to build an accurate mathematical model. Therefore, an effective surface roughness prediction method is of great significance to improve machining efficiency and reduce machining cost. In this paper, taking flank milling Ti6Al4 V alloy as an example, a surface roughness prediction framework (SRPF) based on CLBAS-BP algorithm is proposed. Chaotic Lorentz system and Lévy flight strategy are used to optimize BAS algorithm, which can improve local search ability, solution precision and convergence speed. CLBAS-BP algorithm has higher prediction accuracy than other algorithms, and can predict workpiece surface roughness with various cutting parameters. This study provides the technique foundation for improving the high-precision manufacturing of products.