Abstract

The prediction of software defects was introduced to support development and maintenance activities to improve the software quality by finding errors early in the software development. It facilitates maintenance in terms of effort, time and more importantly the cost prediction for software evolution and maintenance activities. In this paper, we evaluate the quality related attributes in developed software products. The software call graph model is also used for several applications in order to represent and reflect the degree of their complexity in terms of understandability, testability and maintainability efforts. The extracted metrics are investigated for the evaluated applications in correlation with bugs collected from customers bug reports. Those software related bugs are compiled into datasets files to use as an input to a data miner for classification, prediction and association analysis. Finally, the analysis results is evaluated in terms of finding the correlation between software products bugs and call graph based metrics. We find that call graph based metrics are appropriate to detect and predict software defects so that the activities of testing and maintenance stages become easier to estimate or assess after the product delivery.