Magnetic structures-the spatial arrangement of magnetic moments-underpin the diverse functionalities of magnetic materials, yet they have been experimentally determined for only a small fraction of known compounds. Conventional first-principles prediction requires candidate magnetic configurations as input, but the number of possible configurations is in principle infinite. Although the classic Landau-theory based symmetry analysis is widely used to classify magnetic ordering, their direct application to generate input magnetic configurations remains a major challenge due to unavoidable residual degrees of freedom. Here we turn this classification language into a practical predictive framework by resolving these freedoms through a two-step symmetry-aided strategy: an irrep-compatible maximal-magnetic-subgroup selection, followed by the physically motivated high-symmetry-direction construction. This highly efficient scheme is justified within the newly developed oriented spin space group framework. Benchmarked against 1,043 Landau-compatible experimental structures in MAGNDATA, our candidate sets reproduce 90% of the experimentally reported magnetic structures in the four high-symmetry crystal systems with only ~51 candidates per material on average, and subsequent first-principles calculations with material-specific Hubbard U parameters recover the experimental ground state with 83% accuracy. Applying the validated workflow to 19,808 transition-metal compounds, we construct a database of more than 5,000 magnetic materials-nearly twice the size of the experimental MAGNDATA. The database reveals statistical regularities accessible only at scale, including the occurrence rate of magnetism, the prevalence of competing magnetic states, and enables the identification of hundreds of magnetic topological materials, altermagnets, and other emerging magnetic phases and functionalities.