The paper introduces the application of principal component analysis (PCA) to extract and reorganize the principal components from various nuclear mass models. The authors, X. H. Wu and P. W. Zhao, use PCA to identify the major patterns and effects contained in different nuclear mass models, which are then recombined to build new mass models. The study focuses on six widely used nuclear mass models: FRDM2012, HFB17, KTUY05, DIM, RMF, and LDM. The PCA process involves transforming the mass predictions of these models into a set of uncorrelated principal components, ordered by their importance in representing the features of the original models. The first principal component (PC1) captures the bulk properties of nuclei, while the second PC2 reflects deformation properties related to shell effects and odd-even effects. The third and fourth PCs (PC3+PC4) represent the breaking of neutron and proton symmetry energy, and the fifth and sixth PCs (PC5+PC6) show more irregular patterns. The authors find that including more principal components improves the accuracy of the new mass models, with the first four PCs achieving a precision of 519 keV, which is better than the smallest of the original six models (HFB17 at 591 keV). The reliability of the new mass models is further validated through extrapolation, showing that they can accurately predict unknown regions without overfitting. The study concludes that PCA can effectively combine the effects from different theoretical models to improve nuclear mass predictions, providing a novel approach to building more accurate nuclear mass tables.The paper introduces the application of principal component analysis (PCA) to extract and reorganize the principal components from various nuclear mass models. The authors, X. H. Wu and P. W. Zhao, use PCA to identify the major patterns and effects contained in different nuclear mass models, which are then recombined to build new mass models. The study focuses on six widely used nuclear mass models: FRDM2012, HFB17, KTUY05, DIM, RMF, and LDM. The PCA process involves transforming the mass predictions of these models into a set of uncorrelated principal components, ordered by their importance in representing the features of the original models. The first principal component (PC1) captures the bulk properties of nuclei, while the second PC2 reflects deformation properties related to shell effects and odd-even effects. The third and fourth PCs (PC3+PC4) represent the breaking of neutron and proton symmetry energy, and the fifth and sixth PCs (PC5+PC6) show more irregular patterns. The authors find that including more principal components improves the accuracy of the new mass models, with the first four PCs achieving a precision of 519 keV, which is better than the smallest of the original six models (HFB17 at 591 keV). The reliability of the new mass models is further validated through extrapolation, showing that they can accurately predict unknown regions without overfitting. The study concludes that PCA can effectively combine the effects from different theoretical models to improve nuclear mass predictions, providing a novel approach to building more accurate nuclear mass tables.