In software testing field, test case reduction has been a research hotspot for a long time. Some researches currently use the complex relationship between test requirements for the test suites of test case, which can optimize the corresponding test suites on this basis. But the corresponding test case of a single test requirement may be a collection of density distribution in large quantities. This paper does an in-depth research and exploration on how to rationally optimize test case for the corresponding test suits of a single test requirement in the premise of test case. It proposes two classes based on black box testing equivalence partitioning and boundary value analysis strategy. Based on DBSCAN algorithm, it proposes a scientific and reasonable parameter selection method, and improves the adaptation degree and efficiency of algorithm. Combined with optimization algorithm and two strategies, it gets the optimal reduction set of test cases.