The SCD model can effectively describe the changes of the durations in the ultra-high time series，but it fixes the logarithmic form for conditional mean function to avoid the negative conditional duration． This paper weakened the restriction of the non-negativity and proposed a Box-Cox SCD model based on the Box-Cox transformation． This new type of SCD model is more flexible，and it can find the most suitable conditional mean function． However，the flexibility is the cost of the complexity of the estimation of the parameters． This paper presented an MCMC estimation and compared the predictions of the Box-Cox SCD model and the SCD model based on the simulated data generated by the TEACD( 1，1) model and by the empirical data of the IF1012 index futures． The empirical study shows that there is obvious clustering in the price durations，and the value of in the Box-Cox SCD model is obviously different from 0，which implies that the logarithmic form of the conditional mean in the SCD model is not reasonable．