Increasing diversity: Natural language measures for software fault prediction

摘要: While challenging, the ability to predict faulty modules of a program is valuable software project because it can reduce cost development, as well maintenance and evolution. Three language-processing based measures are introduced applied problem fault prediction. The first measure on usage natural language in program's identifiers. second concerns conciseness consistency third measure, referred QALP score, makes use techniques from information retrieval judge quality. score has been shown correlate with human judgments Two case studies consider processing applicability prediction using two programs (one open source, one proprietary). Linear mixed-effects regression models used identify relationships between defects measures. Results, while complex, show that improve prediction, especially when combination. Overall, explain one-third two-thirds faults studies. Consistent other uses processing, value three increases size module considered.