A Theoretical and Empirical Investigation of the Validity of the Current Software Reliability Prediction Techniques

• Main Authors: Ping-Chun Tsai, 蔡秉均
• Other Authors: Chin-FengFan
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/64014259231791482925

碩士 === 元智大學 === 資訊工程學系 === 100 === Reliability is an important quality indicator. It can be used to confirm whether a product could perform its expected function. Howerver, there are many doubts about the concept of software reliablilty. Software reliability predition techniques had developed for nearly 40 years since the 1970's. But nowadays, few software reliability prediction models can be widely accepted by academia and industry. Therefore, in our study, we aimed to resolve the doubts on software reliability, and identify the reasons why software reliability predictive value could be "inaccurate" at some conditions. According to the concept of "failure physics", we proposed a 7-layers reliability prediction model. To investigate why software reliability could sometimes be "inaccurate" in the current methods of software reliability prediction, we applied this model to two cases for experiment. Afterwards, we summarized the results of the experiment, then defined the condition in which the software reliability prediction value would be invalid, and the condition in which the software reliability prediction value would be valid.