The ability of making termination decision is an important index that influences on the longterm performance venture capital firms. Present termination decision methods neglect the information's effect in the process of venture project development. This paper induces and analyses the information during the venture project development process firstly; then chooses the good (bad) signal observed by venture capitalist as the binary learning signal and proposes signal learning model from the perspective of Bayesian posterior estimate considering three conditions that the good (bad) signal observed along with project development. The paper further studies the exogenous optimal termination point according to venture capitalist's altitude to risktaking and finally gives a corresponding example.