Single nucleotide polymorphism (SNP) profiling is a commonly used genetic tool for individual identification and ancestry inference in forensic genetics. This study collected ancestry informative SNPs (AISNPs) from literature and public libraries, and applied softmax regression, support vector machine and random forest, which were used to infer ancestry origins of Northern Han, Japanese and Korean, the three major populations in the North of East Asia. We analyzed 428 AISNPs in 103 northern Han samples and 104 Japanese samples from the 1 000 Genomes Project and 100 Korean samples from the Asian Diversity Project, using multiple linear regression collinearity diagnostics and random forest mean decrease accuracy to screen and optimize high-information AISNPs combinations which were used for ancestry inference linear and nonlinear prediction models, respectively. We constructed two discriminant models of softmax regression and support vector machine with 67-plex AISNPs and a random forest discriminant model with 42-plex AISNPs, achieving high-precision division of Northern Han, Japanese and Korean. The accuracy rates of the 5 times 10-fold cross-validation test of the softmax regression model, support vector machine model and random forest model were 95.19%, 95.77%, and 94.53%, respectively. The 67-plex and 42-plex AISNP prediction models established in this study can be used for genetic inference of the three major populations in the North of East Asia with high practical application value.