Failure is a four-letter word

摘要: Background: The past years have seen a surge of techniques predicting failure-prone locations based on more or less complex metrics. Few these metrics are actionable, though.Aims: This paper explores simple, easy-to-implement method to predict and avoid failures in software systems. IROP links elementary source code features known lightweight, fashion.Method: We sampled the Eclipse data set mapping defects files three releases. used logistic regression associate programmer actions with defects, tested predictive power resulting classifier terms precision recall, isolated most defect-prone actions. also collected initial feedback possible remedies.Results: In our sample set, correctly predicted up 74% modules, which is par elaborate predictors available. four easy-to-remember recommendations, telling programmers precisely what do errors. Initial from developers suggests that recommendations straightforward follow practice.Conclusions: With abundance development data, even simplest methods can produce "actionable" results.