There are many kinds of software vulnerability prediction models which are capable of predicting the total number and the time interval of occurrence of vulnerabilities in the software. But none of them can predict the severity of software vulnerabilities. However

in some cases

such as software credibility

we have to consider the total number of software vulnerabilities and time interval as well as the vulnerability severity affecting the trustworthiness of software. Considering the impact of the vulnerability severity

the application and risk control of software is also very important in the traditional software security. Based on the traditional Markov model

we classified the severity of software vulnerabilities occurrence

proposed a new software vulnerability prediction mathematical model. The model can not only predict the total number of software vulnerability and the time interval

but also the total number vulnerabilities of each class as well as the type of the vulnerabilities. Our tests showed that it has better accuracy

and the type of information that other prediction models can not offer.