Estimation and Improvement of Software Quality using Time Series Data from Software Development Process

点击翻译按钮获取中文摘要

In this research, we tried to develop a quality estimation method for software development which emphasizes on practical use, especially, ease for the implementation in the actual development field. To do so, we consider the following three viewpoints :1.Prediction of software quality suing the metrics data collected from the design and coding phases in the software development.We proposed a method to predict final quality of products using time series data of detected defects in design and coding phases. In more detail, we classified all projects into 4 groups according to the trend of defects detection. We then analyzed resultant quality of the classes.2. Prediction of software quality using subjective data such as validity of development plan and difference between planned and actual progress of software development.Here, in order to improve the accuracy of prediction, we introduced additional data, subjective questionnaire about software risks at the beginning of project. We considered that the software risks deeply related to the final quality of product. We modeled the relationship among risks and final quality using Bayesian belief network.3. Method to improve quality using simulation techniqueFinally we tried to simulate the software risk and to estimate quality of software products for each phase of software development. To do so, we developed a system to estimate software quality using Bayesian classifier at each phase of development. By using this method, practical estimation of software quality can be achieved.