Abstract: By employing the collective behavior theory and the classical threshold model, this paper firstly studies the characteristics of investors’ behaviors in the stock markets, then based on that, it adds two elements of the complex network and the asynchronous time to establish an extended threshold model. Through computer simulation, it studies the dynamical diffusion of stock market risks in the complex networks. The findings show that:（1）firstly, when the characteristics of risk threshold value are activated, no matter what kind of the network structure is, there will be the stock market disasters; secondly, the expansion of the investor scale can increase the threshold value to activate stock market disasters, thus it can reduce the risk of stock market crash; thirdly, the risk volatility presents an interval effect, when the investment preference heterogeneity is highly convergent or divergent, the possibility of risk diffusion is less and the speed is slower in the complex network, the reason is that the complex network is easy to promote the trend more consistent or discrete, so as to reduce the risk of the whole system. However, in the normal securities markets, the investment preference heterogeneity is located at the intermediate state, the information dissemination of the complex network will drive the speed of risk diffusion to be faster, and the risk will be stronger and wider. Therefore, to take advantage of the power of “Internet Acquaintance Society” to promote the decision-making behaviors for still higher consistentance or higher dissociation can help to reduce overall risk of the financial system. This has certain implications for investors and financial system regulation.

Key words: collective behavior; threshold value model; stock market risks; complex network