With the progress of the times, the international financial market link is becoming more and more closely, while the probability of extreme events more and more high, if there are some indicators can be used as a prediction of the crash, as whether to sell the stocks, it can be very useful. The study process for the use of the Fama-French five-factor model, as well as the VaR model, with the cluster analysis method, and clustering for Taiwan 50 constituent stocks in accordance with the five-factor characteristics of the individual stocks, the similar nature of stock into the same group, the establishment of portfolio, the use of portfolio daily returns to calculate the the VaR, and observe the VaR spread before the crash, how the trend, and whether certain characteristics. Comparison of the cluster group for the predictive ability of the collapse events, as well as the relationship between risk factors and predictive ability. The results of VaR spread movements are often subject to fluctuations significantly change the situation before the crash occurs. By intense will be stable or from stable will be severe. Good predictive ability of the cluster, often its constituent stocks and the collapse of the reasons more closely the relationship. Financial stocks sensitive to the financial tsunami; Electronic stocks are subject to exchange rate affect.Overall, the group with the best predictive ability is more sensitive to momentum effects and investor sentiment indicators ,but non-sensitive to book-to-market factor.To use the Var spread as a predictor of reference,choosing to meet the aforementioned conditions of stocks to the portfolio is a nice way.