Building Practical Software Quality Models

点击翻译按钮获取中文摘要

A software total cost of ownership (TCO) includes all the costs to acquire and keep a program running, from specifications to licenses to training. Maintenance contributes between 50% and 70% of TCO. It includes the costs to change the program to fix an old feature and add new features. Companies need means to assess and improve their program maintainability to reduce their TCO. ISO9126 states that four factors impact developers when changing/maintaining a program: its analysability, changeability, stability, and testability; which form the quality characteristic of maintainability and lead to two main problems. Prb1: developers and their team leaders must assess program maintainability to decide whether it is worth changing the program. Prb2: developers need help in identifying different alternative sets of changes that would improve their program maintainability. The field of (empirical) software engineering now provides a wealth of facts and data to build maintainability assessment models for programs. Yet, no previous work put all these together to produce practical maintainability assessment models. Also, beyond assessing maintainability, developers must also identify and weigh changes necessary to improve their program maintainability before fixing/implementing features. They need practical maintainability improvement models providing different alternative sets of changes. My goal is to build a recommender system, based on practical maintainability assessment and improvement models, to quantify and increase maintainability and help developers facing Prb1 and Prb2. I will use systematic literature reviews to understand program maintainability and transformations. I will collect programs and related data pertaining to their maintainability, in particular experts' opinions, and build a database of samples with which to train quality assessment and improvements models. I will evaluate the models empirically and will develop research prototypes to collect data on programs, evaluate their maintainability, and provide suggestions of improvements using the models. The resulting recommander system will help developers and team leaders reduce maintainability costs and, thus, their program TCO.