Defect Prediction in Medical Software Using Hybrid Genetic Optimized Support Vector Machines

In recent years Food and Drug Administration (FDA) has included medical software into its ambit of regulation. It has issued extensive guidance to regulate medical software. Managing and maintaining defects in medical software using automated techniques is an emerging application area. Software defects are faults in software modules leading to application failure. Defects have been successfully predicted using software metrics and bug reports. Organizations prefer defect prediction in software systems before deployment to gauge quality and maintenance effort. In this work, the software defect prediction problem in medical software is addressed. For using Support Vector Machine (SVM) effectively for either classification problem or regression problem on data which are non-linearly separable, the kernel function parameters have to be efficiently chosen. A Hybrid Genetic Algorithm (GA) is proposed for optimizing the SVM classifier. Experiments show that proposed technique can be effectively used for defect prediction leading to better maintenance.