<p>NdFeB powders with tailored morphological attributes were produced by varying jet milling conditions. A suite of critical parameters characterizing the powders, green compacts, and final magnets were meticulously measured. Utilizing this comprehensive experimental dataset, we developed a machine learning (ML) model to rapidly and accurately predict the remanence (<i>B</i>_r) and coercivity (<i>H</i>_cJ) of the magnets. The SHAP (SHapley Additive exPlanations) technique was employed to interpret the model and quantify the contribution of each input feature. Based on the high-throughput predictions of the ML model and microstructure analysis, the influence of powder particle morphology on the microstructure and properties of the magnets was explored. The results show that the powder particle morphology is an important factor, affecting the performance of the magnets as well as the subsequent process and the structure of the magnets. It was found that refining the powder particle size and narrowing its distribution are instrumental in optimizing the grain boundary structure, leading to finer, more uniform main phase grains and a consequent substantial enhancement in <i>H</i>_cJ. However, when the original powder’s particle size is too small, it leads to powder agglomeration and increases magnet oxygen content. The orientation degree, the density and the remanence of the magnets are all high when the sauter mean diameter (SMD) is between 2.400 and 2.700&#xa0;μm and the particle size distribution (<i>D</i>99<i>/D</i>10) is less than 7. The predicted data show that when the SMD is 2.530&#xa0;μm and <i>D</i>99<i>/D</i>10 is 6.20, the <i>B</i>_r and the <i>H</i>_cJ increase by 4.32 and 8.50%, respectively. This study provides a reference for the technical research and development of S-NdFeB magnets with high performance.</p> Graphical Abstract <p></p>