To further improve the accuracy and convenience of prediction on the surface roughness （R_a） of milled 7475 aluminum alloy， the least squares vector machine （LSSVM） was optimized based on the Aquila Optimizer algorithm （AO）. The R_a prediction model of milled aluminum alloy was constructed with four milling parameters as the input values and R_a as the output value. This model was compared with two other algorithms， PSO-LSSVM and LSSVM. The correlation between the milling parameters and surface roughness was analyzed using gray correlation . The R_a prediction system was built through the GUI interface. The results show that the prediction error of the R_a prediction model based on AO-LSSVM is 4.2876%， and the goodness-of-fit reaches 0.93864， which is better than other algorithms. The correlation between feed per tooth and R_a is the largest， with a gray correlation value of 0.764. The GUI prediction application system can realize efficient， convenient， and accurate prediction of R_a values.