Date Added: Nov 2009
Software quality lags far behind the quality of any other engineered product that humans create, and yet software systems are essential to the modern world. Software engineers have long proposed reusing existing software components to improve overall system quality. As part of a reuse process, the reliability of a component must be established; most commonly, this is done by using a software defect-prediction model. However, research in software defect prediction suffers from a dearth of high-quality datasets that link defect rates to modern (object-oriented) software metrics. While some datasets are large (thousands of records), some contain object-oriented metrics, and some link metrics to defect counts, there are no large datasets that link object-oriented metrics to defect counts.