A Quantitative Approach to Software Maintainability Prediction

Abstract

Software maintainability is one important aspect in the evaluation of software evolution of a software product. Due to the complexity of tracking maintenance behaviors, it is difficult to accurately predict the cost and risk of maintenance after delivery of software products. In an attempt to address this issue quantitatively, software maintainability is viewed as an inevitable evolution process driven by maintenance behaviors, given a health index at the time when a software product are delivered. A Hidden Markov Model (HMM) is used to simulate the maintenance behaviors shown as their possible occurrence probabilities. And software metrics is the measurement of the quality of a software product and its measurement results of a product being delivered are combined to form the health index of the product. The health index works as a weight on the process of maintenance behavior over time. When the occurrence probabilities of maintenance behaviors reach certain number which is reckoned as the indication of the deterioration status of a software product, the product can be regarded as being obsolete. Longer the time, better the maintainability would be.