Software metrics play a major role in predicting software quality; they help project managers in decision-making. Indeed, software metrics provide a quantitative approach allowing the control and the improvement of the development process including the maintenance. The ISO/IEC international standard (14598) on software product quality states, “Internal metrics are of little value unless there is evidence that they are related to external quality”. Many empirical prediction models are presented in the literature; their goal is to investigate the relationship between internal metrics and external qualities, in order to assess software quality. In this paper, we explore different machine-learning (ML) algorithms provided by an open source data-mining environment. We analyse their capacities to produce accurate and usable predictive models.