School buildings in Taiwan are designed not only as places of education, but also as temporary shelters in the aftermath of major earthquakes. Therefore, seismic resistance assessments of existing school buildings are very important. Another issue worth discussing is how to choose the appropriate number of school case studies and seismic factors when using artificial intelligence (AI) to infer the seismic resistance of school buildings. This is because when the number is too high, a large amount of time and money will be required to build the seismic database. If the number is too low, then the inference results will be unsatisfactory. This study applied different research methods to address these issues. First, sensitivity analysis was conducted to determine the optimal number of seismic factors and school case studies. Then, gray theory was utilized to explore the correlations between the seismic factors and seismic resistance of school buildings. Finally, a support vector machine (SVM) and gene expression programming (GEP) were used to determine the optimal assessment models. The SVM was also validated by 10-fold cross-validation and back-propagation network. The results show that when researchers used AI theories, at least five factors were considered. If researchers want to obtain a better inference result, then the number of factors can exceed 10. With regard to the number of cases, it should be twice the number of factors, but more than three times is ideal. For the seismic assessment models inferred by SVM and GEP, the root mean square error of the testing cases ranged from 0.0501 to 0.0541, and 0.0757 to 0.0981, respectively, which indicates good performance. The results can be adopted by structural engineers and architects. The research methods developed in this study can also be referenced by researchers in the future.