Stochastic model of software reliability based on exponential smoothing techniques

Software reliability models can provide quantitative measures for evaluating the reliability of software systems that are growing more important today. By analyzing the traditional double exponential model and C-32 exchange software system in detail, the auto-adapted double exponential smoothing model, which the smoothing coefficients can be adjusted in order to adapt the characters of time sequences, was established. Then a kind of dual stochastic model of software reliability based on exponential smoothing also was put forward, which builds the reliability model for software failures by constructing a compensatory stochastic process through exponential smoothing technique. After comparing the predictions from auto-adapted double exponential smoothing model and dual stochastic model based on exponential smoothing with the predictions from traditional exponential smoothing model, G-O model and factual data of software failure, it can be seen that the two improved model not only inherit the fine characters of exponential smoothing models but also has the ability of providing more accurate prediction.