关键词: 软件故障预测, 软件度量, 机器学习, 统计方法

Abstract: By predicting software faults with metrics information， the testing and verification resources are allocated effectively. Software fault static prediction methods are summarized from the point of view of available metrics data and predicting models. Available metrics data has method-level， class-level， component-level， file-level and process-level categories. Predicting models have machine learning and statistics categories. Some issues to be studied are summarized， which are the performance assessing indexes， availability of the metrics and the effect of fault types on the fault prediction.

Key words: software fault prediction, software metrics, machine learning, statistics