This paper summarized theories and methods for the combination of interrelated evidences , discussed the scourge of current combination methods , and presented a new combination method for interrelated evidences through analysis of focal element of interrelated evidences. The method focuses on the adjustment of basic probability num2 bers of interrelated focal element . It improves current adjustment methods that adjust all focal elements , makes the adjustment object more reasonable. The adjustment coefficient can be obtained through learning based optimal meth2 ods. This method can be used to combine interrelated evidences while the basic probability numbers of interrelated evidences are unavailable. An experimental computing shows the results is satisfactory.