Traditional natural products have emerged as a valuable source of drug development due to their suitability to polypharmacology and empirical knowledge about their efficacy and safety. But conventional databases of the traditional natural products have limitations with respect to compatibility and manageability.The objective of this research is to identify multi-component drugs from traditional natural products. For the purpose, we constructed an integrated database of traditional natural products, Material-Efficacy Matrix Database (MEMDB). The database has enhanced compatibility by mapping herbs and molecular components to international identifiers, and improved manageability by structuralizing the functional information of prescriptions and herbs though text-mining methods. Employing the constructed database, we identified multi-component drugs for 12 target phenotypes respectively. Concretely, essential combinations of molecular components were inferred as candidate drugs with an association rule mining method. The results were validated indirectly by evaluating their explanatory power for herbs’ effect on target phenotypes. As a case study, it was shown that the results for diabetes mellitus are consistent with previous studies.