Software Defect Prediction (SDP) is one of the most used activities comes under the Testing Phase of SDLC(software development life cycle). Software defect predictions(SDP) identifies the modules that defect prone and require extensive testing. This way, the testing resources can be used efficiently without disturbing the function of constraints. Software defect predictions are very helpful in testing, it is not always easy to predict the defective modules. There are various issues that define the smooth performance and also the use of the Defect Prediction models. In this article, we have discussed some of the major issues of SDP and studied what has been done so far to address them.
Software systems play an important role in our digital lives, their complexity continues to increase. The increased complexity of software systems makes the importance of their quality very difficult. And that is the reason a significant amount of recent research focuses on the prioritization of software quality assurance efforts. Software defect predictions are made to determine where future defects might appear. Since then, there have been many types of research and many accomplishments in the area of software defect prediction.
Software defect prediction can be done by Supervised and unsupervised learning. Before understanding the impact on supervised learning and unsupervised learning on software defect prediction, first, you need to know about what is supervised learning and what is unsupervised learning?
Supervised learning
Supervised learning is defined as the task of inferring a function from a labeled set of training data. By fitting to the set of the labeled training set, users want to find the most optimal model parameters to predict unknown labels on other objects data set. If the label obtains with the result of a real number, we call the working as a regression technique. If the label is obtained from the limited number of values, where these values are unordered, then it is a classification technique.
Unsupervised learning
In unsupervised learning we need less information about objects, in particular, the train data set is obtained without the label. What is our goal now? It is possible to observe some similarities between groups of objects and include them in appropriate clusters technique. Some objects can differ from all clusters, for this reason, we assume these objects to be available in a group of anomalies.
Software defect predictions with the help of supervised and unsupervised learning.
Software defect prediction has recently attracted the attention of many software quality testers. Some of the major areas in current and different project management software are to effectively utilize resources to make a meaningful impact on time and cost. A programmatic assessment of metrics is essential in order to define the quality of software and to ensure corrective measures. Software defect prediction methods are used to study the impact areas in software using different techniques which consist of a neural network (NN) techniques, clustering techniques, statistical method, and machine learning methods. Data mining are applied in building software defect prediction models which improve the software quality. The performance of three data mining classifier algorithms such as J48, Random Forest, and Naive Bayesian Classifier (NBC) are evaluated based on various criteria like ROC, Precision, MAE, RAE etc. Clustering technique is then applied to the set of data using various technologies such as k-means, Hierarchical Clustering and Make Density-Based Clustering algorithm.