A novel probabilistic-ABC based boosting model for software defect detection

As the development platforms in software technology grow rapidly; the development of software systems has become more and more complex and versatile. It is quite necessary for the software industries to develop software products with good qualities. As the size of the product and development technologies increasing, there is need for identification of software defects at the early stage of development phase. Detection of defects appropriately and accurately measured in the early stage to increase the quality of the product. Software metric can be defined as building blocks of prediction models. The evaluation of software metric is carried out in the design and coding phase prior to actual implementation of the product. Defects are described as a type of conditions that does not satisfy minimum requirement and this can lead to malfunction or unexpected outcomes. The numbers of defects are inversely proportional with the quality which means, on decreasing numbers of defects will result better software quality. Detection of these defects at the early stage of development life cycle will reduce the cost and effort to a great extent. Since decades many researchers have proposed different models for detection of defects in software modules. In this paper we emphasized on defect detection approaches using machine learning scheme to improve true positive rate. Prediction metrics are basically split into two categories, those are: - code metrics and process metrics. In this paper we have designed and implemented a novel defect prediction model using probabilistic ABC based classification model. We compared our model with various defect detection approaches in terms of statistical measures. Experimental results proved that proposed model has high computational detection rate compared to traditional models.