SOFTWARE DEFECT PREDICTION WITH NEWER AND EFFCIENT TECHNIQUES

Abstract

For a long time, software defects have been a big problem in the software optimization and development processes. Various processes have been hindered in the past due to a small defect, which if predicted or corrected at that moment could have resulted in exponential efficiency and enormous commercial benefits. With that in mind it is a viable option of the current time to come up with a process, service or algorithm that can help in software defect prediction. In a SDLC, it is a crucial part in the testing phase to identify the areas that are prone to problem and may ultimately lead to defects which may cause a very massive problem in the later stages of development. These problems may lead to a very important component of the algorithm to malfunction and at that stage can also make it costly to repair in terms of both finance and efforts. Time and cost have both been important development considerations when taking up and developing a project and if it is prolonged after the expected deadlines can lead to massive drop off in efficiency and vulnerability on the execution side of things. It is always better to predict the areas of the problems that may creep in and use algorithms that may help do such process in an easier and effective way. Moreover, we can also use these algorithms with greater and much accurate results when the training has been done for long periods of time as it can make the procedure for training much more robust and precise. This study helps in the analysis of the previously applied activities in the field of defect prediction on a widely known dataset and then focusses to optimize on these techniques using complementary methods that have been widely used in other disciplines. Newer techniques show promising results.