The paper aims to accurately predict the surface quality of the compacted graphite cast iron during machining and effectively guide the adjustment of machining parameters to ensure stable machining quality, applies a support vector machine model based on differential evolution algorithm optimization (DE-SVM), so as to establish a prediction model of surface roughness of compacted graphite cast iron and a selection method of machining parameters. DE-SVM was used to improve the prediction accuracy of support vector machine regression model, and a prediction model of surface roughness (Ra) for specific machining materials was established. On this basis, the relationship between surface roughness and machining parameters was explored to obtain more suitable machining parameters. Combining the milling experiment data of compacted graphite cast iron, the comparison was carried out between DE-SVM and the SVM model optimized by the commonly used optimization algorithms (particle swarm optimization algorithm (PSO) and genetic algorithm (GA)). The values of MAPE (0.1221) and R2 (0.9559) obtained by DE-SVM model were superior to those of the support vector machine model optimized by particle swarm optimization algorithm and genetic algorithm. Within the given machining parameters, the cutting speed and the feed rate had a great influence on Ra, which was directly proportional to cutting speed and inversely proportional to feed rate; and the depth of cut had no significant effect on Ra. The experimental results indicate that the surface roughness model of compacted graphite cast iron based on DE-SVM model has higher prediction accuracy. The influence of machining parameters on surface roughness obtained by DE-SVM can effectively guide the selection and adjustment of machining parameters. It has good guiding significance for maintaining the excellent machining quality of compacted graphite cast iron.