Most　code　clone　detection　approaches　identify　clones　via　static　source　code　analysis.　Such　approaches　are　effective　and　efficient　in　detecting　lexically　similar　clones.　However,　they　are　less　effective　in　detecting　semantic　clones　that　are　similar　in　functionality　but　different　in　implementation.　As　an　initial　try　to　detect　semantic　clones,　in　this　paper,　we　propose　a　test-based　approach　to　detecting　methods　that　are　semantically　equivalent　to　API　methods.　For　a　given　method　m,　we　generate　its　test　cases　automatically　and　search　for　semantically　equivalent　API　methods　by　running　the　generated　test　cases.　If　two　methods　generate　the　same　output　on　each　of　the　test　cases,　they　are　taken　as　semantically　equivalent　methods.　One　of　the　weakness　of　test-based　clone　detection　is　that　it　is　often　time　consuming.　To　reduce　the　time　complexity,　we　take　the　following　measures.　First,　we　focus　on　methods　instead　of　arbitrary　fragments.　Second,　for　a　given　method,　we　only　compare　it　against　such　API　methods　whose　signatures　are　highly　similar　to　that　of　the　given　method.　We　evaluate　the　proposed　approach　on　10　well-known　applications.　Evaluation　results　suggest　that　it　is　efficient　and　accurate,　and　its　precision　is　up　to　98%.